fix(deps): install transitive vertexai dependency (#10328) to release v3.0 (#10334)

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

.github/workflows/deployment.yml

@@ -29,20 +29,32 @@ jobs:
       build-backend-craft: ${{ steps.check.outputs.build-backend-craft }}
       build-model-server: ${{ steps.check.outputs.build-model-server }}
       is-cloud-tag: ${{ steps.check.outputs.is-cloud-tag }}
-      is-stable: ${{ steps.check.outputs.is-stable }}
       is-beta: ${{ steps.check.outputs.is-beta }}
-      is-stable-standalone: ${{ steps.check.outputs.is-stable-standalone }}
       is-beta-standalone: ${{ steps.check.outputs.is-beta-standalone }}
-      is-craft-latest: ${{ steps.check.outputs.is-craft-latest }}
+      is-latest: ${{ steps.check.outputs.is-latest }}
       is-test-run: ${{ steps.check.outputs.is-test-run }}
       sanitized-tag: ${{ steps.check.outputs.sanitized-tag }}
       short-sha: ${{ steps.check.outputs.short-sha }}
     steps:
+      - name: Checkout (for git tags)
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # ratchet:actions/checkout@v6
+        with:
+          persist-credentials: false
+          fetch-depth: 0
+          fetch-tags: true
+
+      - name: Setup uv
+        uses: astral-sh/setup-uv@5a095e7a2014a4212f075830d4f7277575a9d098 # ratchet:astral-sh/setup-uv@v7
+        with:
+          version: "0.9.9"
+          enable-cache: false
+
       - name: Check which components to build and version info
         id: check
         env:
           EVENT_NAME: ${{ github.event_name }}
         run: |
+          set -eo pipefail
           TAG="${GITHUB_REF_NAME}"
           # Sanitize tag name by replacing slashes with hyphens (for Docker tag compatibility)
           SANITIZED_TAG=$(echo "$TAG" | tr '/' '-')
@@ -54,9 +66,8 @@ jobs:
           IS_VERSION_TAG=false
           IS_STABLE=false
           IS_BETA=false
-          IS_STABLE_STANDALONE=false
           IS_BETA_STANDALONE=false
-          IS_CRAFT_LATEST=false
+          IS_LATEST=false
           IS_PROD_TAG=false
           IS_TEST_RUN=false
           BUILD_DESKTOP=false
@@ -67,9 +78,6 @@ jobs:
           BUILD_MODEL_SERVER=true
 
           # Determine tag type based on pattern matching (do regex checks once)
-          if [[ "$TAG" == craft-* ]]; then
-            IS_CRAFT_LATEST=true
-          fi
           if [[ "$TAG" == *cloud* ]]; then
             IS_CLOUD=true
           fi
@@ -97,20 +105,28 @@ jobs:
             fi
           fi
 
-          # Craft-latest builds backend with Craft enabled
-          if [[ "$IS_CRAFT_LATEST" == "true" ]]; then
-            BUILD_BACKEND_CRAFT=true
-            BUILD_BACKEND=false
-          fi
-
           # Standalone version checks (for backend/model-server - version excluding cloud tags)
-          if [[ "$IS_STABLE" == "true" ]] && [[ "$IS_CLOUD" != "true" ]]; then
-            IS_STABLE_STANDALONE=true
-          fi
           if [[ "$IS_BETA" == "true" ]] && [[ "$IS_CLOUD" != "true" ]]; then
             IS_BETA_STANDALONE=true
           fi
 
+          # Determine if this tag should get the "latest" Docker tag.
+          # Only the highest semver stable tag (vX.Y.Z exactly) gets "latest".
+          if [[ "$IS_STABLE" == "true" ]]; then
+            HIGHEST_STABLE=$(uv run --no-sync --with onyx-devtools ods latest-stable-tag) || {
+              echo "::error::Failed to determine highest stable tag via 'ods latest-stable-tag'"
+              exit 1
+            }
+            if [[ "$TAG" == "$HIGHEST_STABLE" ]]; then
+              IS_LATEST=true
+            fi
+          fi
+
+          # Build craft-latest backend alongside the regular latest.
+          if [[ "$IS_LATEST" == "true" ]]; then
+            BUILD_BACKEND_CRAFT=true
+          fi
+
           # Determine if this is a production tag
           # Production tags are: version tags (v1.2.3*) or nightly tags
           if [[ "$IS_VERSION_TAG" == "true" ]] || [[ "$IS_NIGHTLY" == "true" ]]; then
@@ -129,11 +145,9 @@ jobs:
             echo "build-backend-craft=$BUILD_BACKEND_CRAFT"
             echo "build-model-server=$BUILD_MODEL_SERVER"
             echo "is-cloud-tag=$IS_CLOUD"
-            echo "is-stable=$IS_STABLE"
             echo "is-beta=$IS_BETA"
-            echo "is-stable-standalone=$IS_STABLE_STANDALONE"
             echo "is-beta-standalone=$IS_BETA_STANDALONE"
-            echo "is-craft-latest=$IS_CRAFT_LATEST"
+            echo "is-latest=$IS_LATEST"
             echo "is-test-run=$IS_TEST_RUN"
             echo "sanitized-tag=$SANITIZED_TAG"
             echo "short-sha=$SHORT_SHA"
@@ -600,7 +614,8 @@ jobs:
             latest=false
           tags: |
             type=raw,value=${{ needs.determine-builds.outputs.is-test-run == 'true' && format('web-{0}', needs.determine-builds.outputs.sanitized-tag) || github.ref_name }}
-            type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-stable == 'true' && 'latest' || '' }}
+            type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-latest == 'true' && 'latest' || '' }}
+            type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-latest == 'true' && 'craft-latest' || '' }}
             type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && env.EDGE_TAG == 'true' && 'edge' || '' }}
             type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-beta == 'true' && 'beta' || '' }}
 
@@ -1037,7 +1052,7 @@ jobs:
             latest=false
           tags: |
             type=raw,value=${{ needs.determine-builds.outputs.is-test-run == 'true' && format('backend-{0}', needs.determine-builds.outputs.sanitized-tag) || github.ref_name }}
-            type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-stable-standalone == 'true' && 'latest' || '' }}
+            type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-latest == 'true' && 'latest' || '' }}
             type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && env.EDGE_TAG == 'true' && 'edge' || '' }}
             type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-beta-standalone == 'true' && 'beta' || '' }}
 
@@ -1249,8 +1264,6 @@ jobs:
             latest=false
           tags: |
             type=raw,value=craft-latest
-            # TODO: Consider aligning craft-latest tags with regular backend builds (e.g., latest, edge, beta)
-            # to keep tagging strategy consistent across all backend images
 
       - name: Create and push manifest
         env:
@@ -1473,7 +1486,8 @@ jobs:
             latest=false
           tags: |
             type=raw,value=${{ needs.determine-builds.outputs.is-test-run == 'true' && format('model-server-{0}', needs.determine-builds.outputs.sanitized-tag) || github.ref_name }}
-            type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-stable-standalone == 'true' && 'latest' || '' }}
+            type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-latest == 'true' && 'latest' || '' }}
+            type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-latest == 'true' && 'craft-latest' || '' }}
             type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && env.EDGE_TAG == 'true' && 'edge' || '' }}
             type=raw,value=${{ needs.determine-builds.outputs.is-test-run != 'true' && needs.determine-builds.outputs.is-beta-standalone == 'true' && 'beta' || '' }}
 

backend/Dockerfile

@@ -47,6 +47,8 @@ RUN apt-get update && \
         gcc \
         nano \
         vim \
+        # Install procps so kubernetes exec sessions can use ps aux for debugging
+        procps \
         libjemalloc2 \
         && \
     rm -rf /var/lib/apt/lists/* && \

backend/onyx/background/celery/tasks/user_file_processing/tasks.py

@@ -24,6 +24,7 @@ from onyx.configs.app_configs import MANAGED_VESPA
 from onyx.configs.app_configs import VESPA_CLOUD_CERT_PATH
 from onyx.configs.app_configs import VESPA_CLOUD_KEY_PATH
 from onyx.configs.constants import CELERY_GENERIC_BEAT_LOCK_TIMEOUT
+from onyx.configs.constants import CELERY_USER_FILE_DELETE_TASK_EXPIRES
 from onyx.configs.constants import CELERY_USER_FILE_PROCESSING_LOCK_TIMEOUT
 from onyx.configs.constants import CELERY_USER_FILE_PROCESSING_TASK_EXPIRES
 from onyx.configs.constants import CELERY_USER_FILE_PROJECT_SYNC_LOCK_TIMEOUT
@@ -33,6 +34,7 @@ from onyx.configs.constants import OnyxCeleryPriority
 from onyx.configs.constants import OnyxCeleryQueues
 from onyx.configs.constants import OnyxCeleryTask
 from onyx.configs.constants import OnyxRedisLocks
+from onyx.configs.constants import USER_FILE_DELETE_MAX_QUEUE_DEPTH
 from onyx.configs.constants import USER_FILE_PROCESSING_MAX_QUEUE_DEPTH
 from onyx.configs.constants import USER_FILE_PROJECT_SYNC_MAX_QUEUE_DEPTH
 from onyx.connectors.file.connector import LocalFileConnector
@@ -88,6 +90,17 @@ def _user_file_delete_lock_key(user_file_id: str | UUID) -> str:
     return f"{OnyxRedisLocks.USER_FILE_DELETE_LOCK_PREFIX}:{user_file_id}"
 
 
+def _user_file_delete_queued_key(user_file_id: str | UUID) -> str:
+    """Key that exists while a delete_single_user_file task is sitting in the queue.
+
+    The beat generator sets this with a TTL equal to CELERY_USER_FILE_DELETE_TASK_EXPIRES
+    before enqueuing and the worker deletes it as its first action.  This prevents
+    the beat from adding duplicate tasks for files that already have a live task
+    in flight.
+    """
+    return f"{OnyxRedisLocks.USER_FILE_DELETE_QUEUED_PREFIX}:{user_file_id}"
+
+
 def get_user_file_project_sync_queue_depth(celery_app: Celery) -> int:
     redis_celery: Redis = celery_app.broker_connection().channel().client  # type: ignore
     return celery_get_queue_length(
@@ -546,7 +559,23 @@ def process_single_user_file(
     ignore_result=True,
 )
 def check_for_user_file_delete(self: Task, *, tenant_id: str) -> None:
-    """Scan for user files with DELETING status and enqueue per-file tasks."""
+    """Scan for user files with DELETING status and enqueue per-file tasks.
+
+    Three mechanisms prevent queue runaway (mirrors check_user_file_processing):
+
+    1. **Queue depth backpressure** – if the broker queue already has more than
+       USER_FILE_DELETE_MAX_QUEUE_DEPTH items we skip this beat cycle entirely.
+
+    2. **Per-file queued guard** – before enqueuing a task we set a short-lived
+       Redis key (TTL = CELERY_USER_FILE_DELETE_TASK_EXPIRES).  If that key
+       already exists the file already has a live task in the queue, so we skip
+       it.  The worker deletes the key the moment it picks up the task so the
+       next beat cycle can re-enqueue if the file is still DELETING.
+
+    3. **Task expiry** – every enqueued task carries an `expires` value equal to
+       CELERY_USER_FILE_DELETE_TASK_EXPIRES.  If a task is still sitting in
+       the queue after that deadline, Celery discards it without touching the DB.
+    """
     task_logger.info("check_for_user_file_delete - Starting")
     redis_client = get_redis_client(tenant_id=tenant_id)
     lock: RedisLock = redis_client.lock(
@@ -555,8 +584,23 @@ def check_for_user_file_delete(self: Task, *, tenant_id: str) -> None:
     )
     if not lock.acquire(blocking=False):
         return None
+
     enqueued = 0
+    skipped_guard = 0
     try:
+        # --- Protection 1: queue depth backpressure ---
+        # NOTE: must use the broker's Redis client (not redis_client) because
+        # Celery queues live on a separate Redis DB with CELERY_SEPARATOR keys.
+        r_celery: Redis = self.app.broker_connection().channel().client  # type: ignore
+        queue_len = celery_get_queue_length(OnyxCeleryQueues.USER_FILE_DELETE, r_celery)
+        if queue_len > USER_FILE_DELETE_MAX_QUEUE_DEPTH:
+            task_logger.warning(
+                f"check_for_user_file_delete - Queue depth {queue_len} exceeds "
+                f"{USER_FILE_DELETE_MAX_QUEUE_DEPTH}, skipping enqueue for "
+                f"tenant={tenant_id}"
+            )
+            return None
+
         with get_session_with_current_tenant() as db_session:
             user_file_ids = (
                 db_session.execute(
@@ -568,23 +612,40 @@ def check_for_user_file_delete(self: Task, *, tenant_id: str) -> None:
                 .all()
             )
             for user_file_id in user_file_ids:
-                self.app.send_task(
-                    OnyxCeleryTask.DELETE_SINGLE_USER_FILE,
-                    kwargs={"user_file_id": str(user_file_id), "tenant_id": tenant_id},
-                    queue=OnyxCeleryQueues.USER_FILE_DELETE,
-                    priority=OnyxCeleryPriority.HIGH,
+                # --- Protection 2: per-file queued guard ---
+                queued_key = _user_file_delete_queued_key(user_file_id)
+                guard_set = redis_client.set(
+                    queued_key,
+                    1,
+                    ex=CELERY_USER_FILE_DELETE_TASK_EXPIRES,
+                    nx=True,
                 )
+                if not guard_set:
+                    skipped_guard += 1
+                    continue
+
+                # --- Protection 3: task expiry ---
+                try:
+                    self.app.send_task(
+                        OnyxCeleryTask.DELETE_SINGLE_USER_FILE,
+                        kwargs={
+                            "user_file_id": str(user_file_id),
+                            "tenant_id": tenant_id,
+                        },
+                        queue=OnyxCeleryQueues.USER_FILE_DELETE,
+                        priority=OnyxCeleryPriority.HIGH,
+                        expires=CELERY_USER_FILE_DELETE_TASK_EXPIRES,
+                    )
+                except Exception:
+                    redis_client.delete(queued_key)
+                    raise
                 enqueued += 1
-    except Exception as e:
-        task_logger.exception(
-            f"check_for_user_file_delete - Error enqueuing deletes - {e.__class__.__name__}"
-        )
-        return None
     finally:
         if lock.owned():
             lock.release()
+
     task_logger.info(
-        f"check_for_user_file_delete - Enqueued {enqueued} tasks for tenant={tenant_id}"
+        f"check_for_user_file_delete - Enqueued {enqueued} tasks, skipped_guard={skipped_guard} for tenant={tenant_id}"
     )
     return None
 
@@ -602,6 +663,9 @@ def delete_user_file_impl(
     file_lock: RedisLock | None = None
     if redis_locking:
         redis_client = get_redis_client(tenant_id=tenant_id)
+        # Clear the queued guard so the beat can re-enqueue if deletion fails
+        # and the file remains in DELETING status.
+        redis_client.delete(_user_file_delete_queued_key(user_file_id))
         file_lock = redis_client.lock(
             _user_file_delete_lock_key(user_file_id),
             timeout=CELERY_GENERIC_BEAT_LOCK_TIMEOUT,

backend/onyx/configs/app_configs.py

@@ -788,6 +788,29 @@ MAX_FILE_SIZE_BYTES = int(
     os.environ.get("MAX_FILE_SIZE_BYTES") or 2 * 1024 * 1024 * 1024
 )  # 2GB in bytes
 
+# Maximum embedded images allowed in a single file. PDFs (and other formats)
+# with thousands of embedded images can OOM the user-file-processing worker
+# because every image is decoded with PIL and then sent to the vision LLM.
+# Enforced both at upload time (rejects the file) and during extraction
+# (defense-in-depth: caps the number of images materialized).
+#
+# Clamped to >= 0; a negative env value would turn upload validation into
+# always-fail and extraction into always-stop, which is never desired. 0
+# disables image extraction entirely, which is a valid (if aggressive) setting.
+MAX_EMBEDDED_IMAGES_PER_FILE = max(
+    0, int(os.environ.get("MAX_EMBEDDED_IMAGES_PER_FILE") or 500)
+)
+
+# Maximum embedded images allowed across all files in a single upload batch.
+# Protects against the scenario where a user uploads many files that each
+# fall under MAX_EMBEDDED_IMAGES_PER_FILE but aggregate to enough work
+# (serial-ish celery fan-out plus per-image vision-LLM calls) to OOM the
+# worker under concurrency or run up surprise latency/cost. Also clamped
+# to >= 0.
+MAX_EMBEDDED_IMAGES_PER_UPLOAD = max(
+    0, int(os.environ.get("MAX_EMBEDDED_IMAGES_PER_UPLOAD") or 1000)
+)
+
 # Use document summary for contextual rag
 USE_DOCUMENT_SUMMARY = os.environ.get("USE_DOCUMENT_SUMMARY", "true").lower() == "true"
 # Use chunk summary for contextual rag

backend/onyx/configs/constants.py

@@ -177,6 +177,14 @@ USER_FILE_PROJECT_SYNC_MAX_QUEUE_DEPTH = 500
 
 CELERY_USER_FILE_PROJECT_SYNC_LOCK_TIMEOUT = 5 * 60  # 5 minutes (in seconds)
 
+# How long a queued user-file-delete task is valid before workers discard it.
+# Mirrors the processing task expiry to prevent indefinite queue growth when
+# files are stuck in DELETING status and the beat keeps re-enqueuing them.
+CELERY_USER_FILE_DELETE_TASK_EXPIRES = 60  # 1 minute (in seconds)
+
+# Max queue depth before the delete beat stops enqueuing more delete tasks.
+USER_FILE_DELETE_MAX_QUEUE_DEPTH = 500
+
 CELERY_SANDBOX_FILE_SYNC_LOCK_TIMEOUT = 5 * 60  # 5 minutes (in seconds)
 
 DANSWER_REDIS_FUNCTION_LOCK_PREFIX = "da_function_lock:"
@@ -469,6 +477,9 @@ class OnyxRedisLocks:
     USER_FILE_PROJECT_SYNC_QUEUED_PREFIX = "da_lock:user_file_project_sync_queued"
     USER_FILE_DELETE_BEAT_LOCK = "da_lock:check_user_file_delete_beat"
     USER_FILE_DELETE_LOCK_PREFIX = "da_lock:user_file_delete"
+    # Short-lived key set when a delete task is enqueued; cleared when the worker picks it up.
+    # Prevents the beat from re-enqueuing the same file while a delete task is already queued.
+    USER_FILE_DELETE_QUEUED_PREFIX = "da_lock:user_file_delete_queued"
 
     # Release notes
     RELEASE_NOTES_FETCH_LOCK = "da_lock:release_notes_fetch"

backend/onyx/connectors/google_utils/google_utils.py

@@ -157,9 +157,7 @@ def _execute_single_retrieval(
             logger.error(f"Error executing request: {e}")
             raise e
         elif _is_rate_limit_error(e):
-            results = _execute_with_retry(
-                lambda: retrieval_function(**request_kwargs).execute()
-            )
+            results = _execute_with_retry(retrieval_function(**request_kwargs))
         elif e.resp.status == 404 or e.resp.status == 403:
             if continue_on_404_or_403:
                 logger.debug(f"Error executing request: {e}")

backend/onyx/db/index_attempt.py

@@ -583,6 +583,67 @@ def get_latest_index_attempt_for_cc_pair_id(
     return db_session.execute(stmt).scalar_one_or_none()
 
 
+def get_latest_successful_index_attempt_for_cc_pair_id(
+    db_session: Session,
+    connector_credential_pair_id: int,
+    secondary_index: bool = False,
+) -> IndexAttempt | None:
+    """Returns the most recent successful index attempt for the given cc pair,
+    filtered to the current (or future) search settings.
+    Uses MAX(id) semantics to match get_latest_index_attempts_by_status."""
+    status = IndexModelStatus.FUTURE if secondary_index else IndexModelStatus.PRESENT
+    stmt = (
+        select(IndexAttempt)
+        .where(
+            IndexAttempt.connector_credential_pair_id == connector_credential_pair_id,
+            IndexAttempt.status.in_(
+                [IndexingStatus.SUCCESS, IndexingStatus.COMPLETED_WITH_ERRORS]
+            ),
+        )
+        .join(SearchSettings)
+        .where(SearchSettings.status == status)
+        .order_by(desc(IndexAttempt.id))
+        .limit(1)
+    )
+    return db_session.execute(stmt).scalar_one_or_none()
+
+
+def get_latest_successful_index_attempts_parallel(
+    secondary_index: bool = False,
+) -> Sequence[IndexAttempt]:
+    """Batch version: returns the latest successful index attempt per cc pair.
+    Covers both SUCCESS and COMPLETED_WITH_ERRORS (matching is_successful())."""
+    model_status = (
+        IndexModelStatus.FUTURE if secondary_index else IndexModelStatus.PRESENT
+    )
+    with get_session_with_current_tenant() as db_session:
+        latest_ids = (
+            select(
+                IndexAttempt.connector_credential_pair_id,
+                func.max(IndexAttempt.id).label("max_id"),
+            )
+            .join(SearchSettings, IndexAttempt.search_settings_id == SearchSettings.id)
+            .where(
+                SearchSettings.status == model_status,
+                IndexAttempt.status.in_(
+                    [IndexingStatus.SUCCESS, IndexingStatus.COMPLETED_WITH_ERRORS]
+                ),
+            )
+            .group_by(IndexAttempt.connector_credential_pair_id)
+            .subquery()
+        )
+
+        stmt = select(IndexAttempt).join(
+            latest_ids,
+            (
+                IndexAttempt.connector_credential_pair_id
+                == latest_ids.c.connector_credential_pair_id
+            )
+            & (IndexAttempt.id == latest_ids.c.max_id),
+        )
+        return db_session.execute(stmt).scalars().all()
+
+
 def count_index_attempts_for_cc_pair(
     db_session: Session,
     cc_pair_id: int,

backend/onyx/db/projects.py

@@ -12,6 +12,7 @@ from sqlalchemy.orm import Session
 from starlette.background import BackgroundTasks
 
 from onyx.configs.app_configs import DISABLE_VECTOR_DB
+from onyx.configs.constants import CELERY_USER_FILE_PROCESSING_TASK_EXPIRES
 from onyx.configs.constants import FileOrigin
 from onyx.configs.constants import OnyxCeleryPriority
 from onyx.configs.constants import OnyxCeleryQueues
@@ -144,6 +145,7 @@ def upload_files_to_user_files_with_indexing(
                 kwargs={"user_file_id": user_file.id, "tenant_id": tenant_id},
                 queue=OnyxCeleryQueues.USER_FILE_PROCESSING,
                 priority=OnyxCeleryPriority.HIGH,
+                expires=CELERY_USER_FILE_PROCESSING_TASK_EXPIRES,
             )
             logger.info(
                 f"Triggered indexing for user_file_id={user_file.id} "

backend/onyx/document_index/vespa/chunk_retrieval.py

@@ -503,20 +503,31 @@ def query_vespa(
             response = http_client.post(SEARCH_ENDPOINT, json=params)
             response.raise_for_status()
     except httpx.HTTPError as e:
-        error_base = "Failed to query Vespa"
-        logger.error(
-            f"{error_base}:\n"
-            f"Request URL: {e.request.url}\n"
-            f"Request Headers: {e.request.headers}\n"
-            f"Request Payload: {params}\n"
-            f"Exception: {str(e)}"
-            + (
-                f"\nResponse: {e.response.text}"
-                if isinstance(e, httpx.HTTPStatusError)
-                else ""
-            )
+        response_text = (
+            e.response.text if isinstance(e, httpx.HTTPStatusError) else None
         )
-        raise httpx.HTTPError(error_base) from e
+        status_code = (
+            e.response.status_code if isinstance(e, httpx.HTTPStatusError) else None
+        )
+        yql_value = params.get("yql", "")
+        yql_length = len(str(yql_value))
+
+        # Log each detail on its own line so log collectors capture them
+        # as separate entries rather than truncating a single multiline msg
+        logger.error(
+            f"Failed to query Vespa | "
+            f"status={status_code} | "
+            f"yql_length={yql_length} | "
+            f"exception={str(e)}"
+        )
+        if response_text:
+            logger.error(f"Vespa error response: {response_text[:1000]}")
+        logger.error(f"Vespa request URL: {e.request.url}")
+
+        # Re-raise with diagnostics so callers see what actually went wrong
+        raise httpx.HTTPError(
+            f"Failed to query Vespa (status={status_code}, " f"yql_length={yql_length})"
+        ) from e
 
     response_json: dict[str, Any] = response.json()
 

backend/onyx/document_index/vespa/shared_utils/vespa_request_builders.py

@@ -43,6 +43,22 @@ def build_vespa_filters(
             return ""
         return f"({' or '.join(eq_elems)})"
 
+    def _build_weighted_set_filter(key: str, vals: list[str] | None) -> str:
+        """Build a Vespa weightedSet filter for large value lists.
+
+        Uses Vespa's native weightedSet() operator instead of OR-chained
+        'contains' clauses.  This is critical for fields like
+        access_control_list where a single user may have tens of thousands
+        of ACL entries — OR clauses at that scale cause Vespa to reject
+        the query with HTTP 400."""
+        if not key or not vals:
+            return ""
+        filtered = [val for val in vals if val]
+        if not filtered:
+            return ""
+        items = ", ".join(f'"{val}":1' for val in filtered)
+        return f"weightedSet({key}, {{{items}}})"
+
     def _build_int_or_filters(key: str, vals: list[int] | None) -> str:
         """For an integer field filter.
         Returns a bare clause or ""."""
@@ -157,11 +173,16 @@ def build_vespa_filters(
     if filters.tenant_id and MULTI_TENANT:
         filter_parts.append(build_tenant_id_filter(filters.tenant_id))
 
-    # ACL filters
+    # ACL filters — use weightedSet for efficient matching against the
+    # access_control_list weightedset<string> field.  OR-chaining thousands
+    # of 'contains' clauses causes Vespa to reject the query (HTTP 400)
+    # for users with large numbers of external permission groups.
     if filters.access_control_list is not None:
         _append(
             filter_parts,
-            _build_or_filters(ACCESS_CONTROL_LIST, filters.access_control_list),
+            _build_weighted_set_filter(
+                ACCESS_CONTROL_LIST, filters.access_control_list
+            ),
         )
 
     # Source type filters

backend/onyx/file_processing/extract_file_text.py

@@ -21,6 +21,7 @@ import chardet
 import openpyxl
 from PIL import Image
 
+from onyx.configs.app_configs import MAX_EMBEDDED_IMAGES_PER_FILE
 from onyx.configs.constants import ONYX_METADATA_FILENAME
 from onyx.configs.llm_configs import get_image_extraction_and_analysis_enabled
 from onyx.file_processing.file_types import OnyxFileExtensions
@@ -176,6 +177,56 @@ def read_text_file(
     return file_content_raw, metadata
 
 
+def count_pdf_embedded_images(file: IO[Any], cap: int) -> int:
+    """Return the number of embedded images in a PDF, short-circuiting at cap+1.
+
+    Used to reject PDFs whose image count would OOM the user-file-processing
+    worker during indexing. Returns a value > cap as a sentinel once the count
+    exceeds the cap, so callers do not iterate thousands of image objects just
+    to report a number. Returns 0 if the PDF cannot be parsed.
+
+    Owner-password-only PDFs (permission restrictions but no open password) are
+    counted normally — they decrypt with an empty string. Truly password-locked
+    PDFs are skipped (return 0) since we can't inspect them; the caller should
+    ensure the password-protected check runs first.
+
+    Always restores the file pointer to its original position before returning.
+    """
+    from pypdf import PdfReader
+
+    try:
+        start_pos = file.tell()
+    except Exception:
+        start_pos = None
+    try:
+        if start_pos is not None:
+            file.seek(0)
+        reader = PdfReader(file)
+        if reader.is_encrypted:
+            # Try empty password first (owner-password-only PDFs); give up if that fails.
+            try:
+                if reader.decrypt("") == 0:
+                    return 0
+            except Exception:
+                return 0
+        count = 0
+        for page in reader.pages:
+            for _ in page.images:
+                count += 1
+                if count > cap:
+                    return count
+        return count
+    except Exception:
+        logger.warning("Failed to count embedded images in PDF", exc_info=True)
+        return 0
+    finally:
+        if start_pos is not None:
+            try:
+                file.seek(start_pos)
+            except Exception:
+                pass
+
+
 def pdf_to_text(file: IO[Any], pdf_pass: str | None = None) -> str:
     """
     Extract text from a PDF. For embedded images, a more complex approach is needed.
@@ -231,8 +282,27 @@ def read_pdf_file(
         )
 
         if extract_images:
+            image_cap = MAX_EMBEDDED_IMAGES_PER_FILE
+            images_processed = 0
+            cap_reached = False
             for page_num, page in enumerate(pdf_reader.pages):
+                if cap_reached:
+                    break
                 for image_file_object in page.images:
+                    if images_processed >= image_cap:
+                        # Defense-in-depth backstop. Upload-time validation
+                        # should have rejected files exceeding the cap, but
+                        # we also break here so a single oversized file can
+                        # never pin a worker.
+                        logger.warning(
+                            "PDF embedded image cap reached (%d). "
+                            "Skipping remaining images on page %d and beyond.",
+                            image_cap,
+                            page_num + 1,
+                        )
+                        cap_reached = True
+                        break
+
                     image = Image.open(io.BytesIO(image_file_object.data))
                     img_byte_arr = io.BytesIO()
                     image.save(img_byte_arr, format=image.format)
@@ -245,6 +315,7 @@ def read_pdf_file(
                         image_callback(img_bytes, image_name)
                     else:
                         extracted_images.append((img_bytes, image_name))
+                    images_processed += 1
 
         return text, metadata, extracted_images
 

backend/onyx/file_processing/file_types.py

@@ -19,12 +19,16 @@ class OnyxMimeTypes:
         PLAIN_TEXT_MIME_TYPE,
         "text/markdown",
         "text/x-markdown",
+        "text/x-log",
         "text/x-config",
         "text/tab-separated-values",
         "application/json",
         "application/xml",
         "text/xml",
         "application/x-yaml",
+        "application/yaml",
+        "text/yaml",
+        "text/x-yaml",
     }
     DOCUMENT_MIME_TYPES = {
         PDF_MIME_TYPE,

backend/onyx/file_processing/image_summarization.py

@@ -88,9 +88,13 @@ def summarize_image_with_error_handling(
     try:
         return summarize_image_pipeline(llm, image_data, user_prompt, system_prompt)
     except UnsupportedImageFormatError:
+        magic_hex = image_data[:8].hex() if image_data else "empty"
         logger.info(
-            "Skipping image summarization due to unsupported MIME type for %s",
+            "Skipping image summarization due to unsupported MIME type "
+            "for %s (magic_bytes=%s, size=%d bytes)",
             context_name,
+            magic_hex,
+            len(image_data),
         )
         return None
 
@@ -134,9 +138,23 @@ def _summarize_image(
         return summary
 
     except Exception as e:
-        error_msg = f"Summarization failed. Messages: {messages}"
-        error_msg = error_msg[:1024]
-        raise ValueError(error_msg) from e
+        # Extract structured details from LiteLLM exceptions when available,
+        # rather than dumping the full messages payload (which contains base64
+        # image data and produces enormous, unreadable error logs).
+        str_e = str(e)
+        if len(str_e) > 512:
+            str_e = str_e[:512] + "... (truncated)"
+        parts = [f"Summarization failed: {type(e).__name__}: {str_e}"]
+        status_code = getattr(e, "status_code", None)
+        llm_provider = getattr(e, "llm_provider", None)
+        model = getattr(e, "model", None)
+        if status_code is not None:
+            parts.append(f"status_code={status_code}")
+        if llm_provider is not None:
+            parts.append(f"llm_provider={llm_provider}")
+        if model is not None:
+            parts.append(f"model={model}")
+        raise ValueError(" | ".join(parts)) from e
 
 
 def _encode_image_for_llm_prompt(image_data: bytes) -> str:

backend/onyx/llm/model_metadata_enrichments.json

@@ -1516,6 +1516,10 @@
     "display_name": "Claude Opus 4.6",
     "model_vendor": "anthropic"
   },
+  "claude-opus-4-7": {
+    "display_name": "Claude Opus 4.7",
+    "model_vendor": "anthropic"
+  },
   "claude-opus-4-5-20251101": {
     "display_name": "Claude Opus 4.5",
     "model_vendor": "anthropic",

backend/onyx/llm/models.py

@@ -46,6 +46,15 @@ ANTHROPIC_REASONING_EFFORT_BUDGET: dict[ReasoningEffort, int] = {
     ReasoningEffort.HIGH: 4096,
 }
 
+# Newer Anthropic models (Claude Opus 4.7+) use adaptive thinking with
+# output_config.effort instead of thinking.type.enabled + budget_tokens.
+ANTHROPIC_ADAPTIVE_REASONING_EFFORT: dict[ReasoningEffort, str] = {
+    ReasoningEffort.AUTO: "medium",
+    ReasoningEffort.LOW: "low",
+    ReasoningEffort.MEDIUM: "medium",
+    ReasoningEffort.HIGH: "high",
+}
+
 
 # Content part structures for multimodal messages
 # The classes in this mirror the OpenAI Chat Completions message types and work well with routers like LiteLLM

backend/onyx/llm/multi_llm.py

@@ -23,6 +23,7 @@ from onyx.llm.interfaces import ToolChoiceOptions
 from onyx.llm.model_response import ModelResponse
 from onyx.llm.model_response import ModelResponseStream
 from onyx.llm.model_response import Usage
+from onyx.llm.models import ANTHROPIC_ADAPTIVE_REASONING_EFFORT
 from onyx.llm.models import ANTHROPIC_REASONING_EFFORT_BUDGET
 from onyx.llm.models import OPENAI_REASONING_EFFORT
 from onyx.llm.request_context import get_llm_mock_response
@@ -67,8 +68,13 @@ STANDARD_MAX_TOKENS_KWARG = "max_completion_tokens"
 _VERTEX_ANTHROPIC_MODELS_REJECTING_OUTPUT_CONFIG = (
     "claude-opus-4-5",
     "claude-opus-4-6",
+    "claude-opus-4-7",
 )
 
+# Anthropic models that require the adaptive thinking API (thinking.type.adaptive
+# + output_config.effort) instead of the legacy thinking.type.enabled + budget_tokens.
+_ANTHROPIC_ADAPTIVE_THINKING_MODELS = ("claude-opus-4-7",)
+
 
 class LLMTimeoutError(Exception):
     """
@@ -193,6 +199,29 @@ def _is_vertex_model_rejecting_output_config(model_name: str) -> bool:
     )
 
 
+def _prompt_contains_tool_call_history(prompt: LanguageModelInput) -> bool:
+    """Check if the prompt contains any assistant messages with tool_calls.
+
+    When Anthropic's extended thinking is enabled, the API requires every
+    assistant message to start with a thinking block before any tool_use
+    blocks.  Since we don't preserve thinking_blocks (they carry
+    cryptographic signatures that can't be reconstructed), we must skip
+    the thinking param whenever history contains prior tool-calling turns.
+    """
+    from onyx.llm.models import AssistantMessage
+
+    msgs = prompt if isinstance(prompt, list) else [prompt]
+    return any(isinstance(msg, AssistantMessage) and msg.tool_calls for msg in msgs)
+
+
+def _anthropic_uses_adaptive_thinking(model_name: str) -> bool:
+    normalized_model_name = model_name.lower()
+    return any(
+        adaptive_model in normalized_model_name
+        for adaptive_model in _ANTHROPIC_ADAPTIVE_THINKING_MODELS
+    )
+
+
 class LitellmLLM(LLM):
     """Uses Litellm library to allow easy configuration to use a multitude of LLMs
     See https://python.langchain.com/docs/integrations/chat/litellm"""
@@ -445,22 +474,42 @@ class LitellmLLM(LLM):
                     }
 
             elif is_claude_model:
-                budget_tokens: int | None = ANTHROPIC_REASONING_EFFORT_BUDGET.get(
-                    reasoning_effort
-                )
+                # Anthropic requires every assistant message with tool_use
+                # blocks to start with a thinking block that carries a
+                # cryptographic signature.  We don't preserve those blocks
+                # across turns, so skip thinking when the history already
+                # contains tool-calling assistant messages.  LiteLLM's
+                # modify_params workaround doesn't cover all providers
+                # (notably Bedrock).
+                has_tool_call_history = _prompt_contains_tool_call_history(prompt)
 
-                if budget_tokens is not None:
-                    if max_tokens is not None:
-                        # Anthropic has a weird rule where max token has to be at least as much as budget tokens if set
-                        # and the minimum budget tokens is 1024
-                        # Will note that overwriting a developer set max tokens is not ideal but is the best we can do for now
-                        # It is better to allow the LLM to output more reasoning tokens even if it results in a fairly small tool
-                        # call as compared to reducing the budget for reasoning.
-                        max_tokens = max(budget_tokens + 1, max_tokens)
-                    optional_kwargs["thinking"] = {
-                        "type": "enabled",
-                        "budget_tokens": budget_tokens,
-                    }
+                if _anthropic_uses_adaptive_thinking(self.config.model_name):
+                    # Newer Anthropic models (Claude Opus 4.7+) reject
+                    # thinking.type.enabled — they require the adaptive
+                    # thinking config with output_config.effort.
+                    if not has_tool_call_history:
+                        optional_kwargs["thinking"] = {"type": "adaptive"}
+                        optional_kwargs["output_config"] = {
+                            "effort": ANTHROPIC_ADAPTIVE_REASONING_EFFORT[
+                                reasoning_effort
+                            ],
+                        }
+                else:
+                    budget_tokens: int | None = ANTHROPIC_REASONING_EFFORT_BUDGET.get(
+                        reasoning_effort
+                    )
+                    if budget_tokens is not None and not has_tool_call_history:
+                        if max_tokens is not None:
+                            # Anthropic has a weird rule where max token has to be at least as much as budget tokens if set
+                            # and the minimum budget tokens is 1024
+                            # Will note that overwriting a developer set max tokens is not ideal but is the best we can do for now
+                            # It is better to allow the LLM to output more reasoning tokens even if it results in a fairly small tool
+                            # call as compared to reducing the budget for reasoning.
+                            max_tokens = max(budget_tokens + 1, max_tokens)
+                        optional_kwargs["thinking"] = {
+                            "type": "enabled",
+                            "budget_tokens": budget_tokens,
+                        }
 
                 # LiteLLM just does some mapping like this anyway but is incomplete for Anthropic
                 optional_kwargs.pop("reasoning_effort", None)

backend/onyx/llm/well_known_providers/recommended-models.json

@@ -1,6 +1,6 @@
 {
-  "version": "1.1",
-  "updated_at": "2026-03-05T00:00:00Z",
+  "version": "1.2",
+  "updated_at": "2026-04-16T00:00:00Z",
   "providers": {
     "openai": {
       "default_model": { "name": "gpt-5.4" },
@@ -10,8 +10,12 @@
       ]
     },
     "anthropic": {
-      "default_model": "claude-opus-4-6",
+      "default_model": "claude-opus-4-7",
       "additional_visible_models": [
+        {
+          "name": "claude-opus-4-7",
+          "display_name": "Claude Opus 4.7"
+        },
         {
           "name": "claude-opus-4-6",
           "display_name": "Claude Opus 4.6"

backend/onyx/server/documents/cc_pair.py

@@ -43,6 +43,9 @@ from onyx.db.index_attempt import count_index_attempt_errors_for_cc_pair
 from onyx.db.index_attempt import count_index_attempts_for_cc_pair
 from onyx.db.index_attempt import get_index_attempt_errors_for_cc_pair
 from onyx.db.index_attempt import get_latest_index_attempt_for_cc_pair_id
+from onyx.db.index_attempt import (
+    get_latest_successful_index_attempt_for_cc_pair_id,
+)
 from onyx.db.index_attempt import get_paginated_index_attempts_for_cc_pair_id
 from onyx.db.indexing_coordination import IndexingCoordination
 from onyx.db.models import IndexAttempt
@@ -190,6 +193,11 @@ def get_cc_pair_full_info(
         only_finished=False,
     )
 
+    latest_successful_attempt = get_latest_successful_index_attempt_for_cc_pair_id(
+        db_session=db_session,
+        connector_credential_pair_id=cc_pair_id,
+    )
+
     # Get latest permission sync attempt for status
     latest_permission_sync_attempt = None
     if cc_pair.access_type == AccessType.SYNC:
@@ -207,6 +215,11 @@ def get_cc_pair_full_info(
             cc_pair_id=cc_pair_id,
         ),
         last_index_attempt=latest_attempt,
+        last_successful_index_time=(
+            latest_successful_attempt.time_started
+            if latest_successful_attempt
+            else None
+        ),
         latest_deletion_attempt=get_deletion_attempt_snapshot(
             connector_id=cc_pair.connector_id,
             credential_id=cc_pair.credential_id,

backend/onyx/server/documents/connector.py

@@ -3,6 +3,7 @@ import math
 import mimetypes
 import os
 import zipfile
+from datetime import datetime
 from io import BytesIO
 from typing import Any
 from typing import cast
@@ -109,6 +110,9 @@ from onyx.db.federated import fetch_all_federated_connectors_parallel
 from onyx.db.index_attempt import get_index_attempts_for_cc_pair
 from onyx.db.index_attempt import get_latest_index_attempts_by_status
 from onyx.db.index_attempt import get_latest_index_attempts_parallel
+from onyx.db.index_attempt import (
+    get_latest_successful_index_attempts_parallel,
+)
 from onyx.db.models import ConnectorCredentialPair
 from onyx.db.models import FederatedConnector
 from onyx.db.models import IndexAttempt
@@ -479,7 +483,9 @@ def is_zip_file(file: UploadFile) -> bool:
 
 
 def upload_files(
-    files: list[UploadFile], file_origin: FileOrigin = FileOrigin.CONNECTOR
+    files: list[UploadFile],
+    file_origin: FileOrigin = FileOrigin.CONNECTOR,
+    unzip: bool = True,
 ) -> FileUploadResponse:
 
     # Skip directories and known macOS metadata entries
@@ -502,31 +508,46 @@ def upload_files(
                 if seen_zip:
                     raise HTTPException(status_code=400, detail=SEEN_ZIP_DETAIL)
                 seen_zip = True
+
+                # Validate the zip by opening it (catches corrupt/non-zip files)
                 with zipfile.ZipFile(file.file, "r") as zf:
-                    zip_metadata_file_id = save_zip_metadata_to_file_store(
-                        zf, file_store
-                    )
-                    for file_info in zf.namelist():
-                        if zf.getinfo(file_info).is_dir():
-                            continue
-
-                        if not should_process_file(file_info):
-                            continue
-
-                        sub_file_bytes = zf.read(file_info)
-
-                        mime_type, __ = mimetypes.guess_type(file_info)
-                        if mime_type is None:
-                            mime_type = "application/octet-stream"
-
-                        file_id = file_store.save_file(
-                            content=BytesIO(sub_file_bytes),
-                            display_name=os.path.basename(file_info),
-                            file_origin=file_origin,
-                            file_type=mime_type,
+                    if unzip:
+                        zip_metadata_file_id = save_zip_metadata_to_file_store(
+                            zf, file_store
                         )
-                        deduped_file_paths.append(file_id)
-                        deduped_file_names.append(os.path.basename(file_info))
+                        for file_info in zf.namelist():
+                            if zf.getinfo(file_info).is_dir():
+                                continue
+
+                            if not should_process_file(file_info):
+                                continue
+
+                            sub_file_bytes = zf.read(file_info)
+
+                            mime_type, __ = mimetypes.guess_type(file_info)
+                            if mime_type is None:
+                                mime_type = "application/octet-stream"
+
+                            file_id = file_store.save_file(
+                                content=BytesIO(sub_file_bytes),
+                                display_name=os.path.basename(file_info),
+                                file_origin=file_origin,
+                                file_type=mime_type,
+                            )
+                            deduped_file_paths.append(file_id)
+                            deduped_file_names.append(os.path.basename(file_info))
+                        continue
+
+                # Store the zip as-is (unzip=False)
+                file.file.seek(0)
+                file_id = file_store.save_file(
+                    content=file.file,
+                    display_name=file.filename,
+                    file_origin=file_origin,
+                    file_type=file.content_type or "application/zip",
+                )
+                deduped_file_paths.append(file_id)
+                deduped_file_names.append(file.filename)
                 continue
 
             # Since we can't render docx files in the UI,
@@ -613,9 +634,10 @@ def _fetch_and_check_file_connector_cc_pair_permissions(
 @router.post("/admin/connector/file/upload", tags=PUBLIC_API_TAGS)
 def upload_files_api(
     files: list[UploadFile],
+    unzip: bool = True,
     _: User = Depends(current_curator_or_admin_user),
 ) -> FileUploadResponse:
-    return upload_files(files, FileOrigin.OTHER)
+    return upload_files(files, FileOrigin.OTHER, unzip=unzip)
 
 
 @router.get("/admin/connector/{connector_id}/files", tags=PUBLIC_API_TAGS)
@@ -1140,21 +1162,26 @@ def get_connector_indexing_status(
             ),
             (),
         ),
+        # Get most recent successful index attempts
+        (
+            lambda: get_latest_successful_index_attempts_parallel(
+                request.secondary_index,
+            ),
+            (),
+        ),
     ]
 
     if user and user.role == UserRole.ADMIN:
-        # For Admin users, we already got all the cc pair in editable_cc_pairs
-        # its not needed to get them again
         (
             editable_cc_pairs,
             federated_connectors,
             latest_index_attempts,
             latest_finished_index_attempts,
+            latest_successful_index_attempts,
         ) = run_functions_tuples_in_parallel(parallel_functions)
         non_editable_cc_pairs = []
     else:
         parallel_functions.append(
-            # Get non-editable connector/credential pairs
             (
                 lambda: get_connector_credential_pairs_for_user_parallel(
                     user, False, None, True, True, False, True, request.source
@@ -1168,6 +1195,7 @@ def get_connector_indexing_status(
             federated_connectors,
             latest_index_attempts,
             latest_finished_index_attempts,
+            latest_successful_index_attempts,
             non_editable_cc_pairs,
         ) = run_functions_tuples_in_parallel(parallel_functions)
 
@@ -1179,6 +1207,9 @@ def get_connector_indexing_status(
     latest_finished_index_attempts = cast(
         list[IndexAttempt], latest_finished_index_attempts
     )
+    latest_successful_index_attempts = cast(
+        list[IndexAttempt], latest_successful_index_attempts
+    )
 
     document_count_info = get_document_counts_for_all_cc_pairs(db_session)
 
@@ -1188,42 +1219,48 @@ def get_connector_indexing_status(
         for connector_id, credential_id, cnt in document_count_info
     }
 
-    cc_pair_to_latest_index_attempt: dict[tuple[int, int], IndexAttempt] = {
-        (
-            attempt.connector_credential_pair.connector_id,
-            attempt.connector_credential_pair.credential_id,
-        ): attempt
-        for attempt in latest_index_attempts
-    }
+    def _attempt_lookup(
+        attempts: list[IndexAttempt],
+    ) -> dict[int, IndexAttempt]:
+        return {attempt.connector_credential_pair_id: attempt for attempt in attempts}
 
-    cc_pair_to_latest_finished_index_attempt: dict[tuple[int, int], IndexAttempt] = {
-        (
-            attempt.connector_credential_pair.connector_id,
-            attempt.connector_credential_pair.credential_id,
-        ): attempt
-        for attempt in latest_finished_index_attempts
-    }
+    cc_pair_to_latest_index_attempt = _attempt_lookup(latest_index_attempts)
+    cc_pair_to_latest_finished_index_attempt = _attempt_lookup(
+        latest_finished_index_attempts
+    )
+    cc_pair_to_latest_successful_index_attempt = _attempt_lookup(
+        latest_successful_index_attempts
+    )
 
     def build_connector_indexing_status(
         cc_pair: ConnectorCredentialPair,
         is_editable: bool,
     ) -> ConnectorIndexingStatusLite | None:
-        # TODO remove this to enable ingestion API
         if cc_pair.name == "DefaultCCPair":
             return None
 
-        latest_attempt = cc_pair_to_latest_index_attempt.get(
-            (cc_pair.connector_id, cc_pair.credential_id)
-        )
+        latest_attempt = cc_pair_to_latest_index_attempt.get(cc_pair.id)
         latest_finished_attempt = cc_pair_to_latest_finished_index_attempt.get(
-            (cc_pair.connector_id, cc_pair.credential_id)
+            cc_pair.id
+        )
+        latest_successful_attempt = cc_pair_to_latest_successful_index_attempt.get(
+            cc_pair.id
         )
         doc_count = cc_pair_to_document_cnt.get(
             (cc_pair.connector_id, cc_pair.credential_id), 0
         )
 
         return _get_connector_indexing_status_lite(
-            cc_pair, latest_attempt, latest_finished_attempt, is_editable, doc_count
+            cc_pair,
+            latest_attempt,
+            latest_finished_attempt,
+            (
+                latest_successful_attempt.time_started
+                if latest_successful_attempt
+                else None
+            ),
+            is_editable,
+            doc_count,
         )
 
     # Process editable cc_pairs
@@ -1384,6 +1421,7 @@ def _get_connector_indexing_status_lite(
     cc_pair: ConnectorCredentialPair,
     latest_index_attempt: IndexAttempt | None,
     latest_finished_index_attempt: IndexAttempt | None,
+    last_successful_index_time: datetime | None,
     is_editable: bool,
     document_cnt: int,
 ) -> ConnectorIndexingStatusLite | None:
@@ -1417,7 +1455,7 @@ def _get_connector_indexing_status_lite(
             else None
         ),
         last_status=latest_index_attempt.status if latest_index_attempt else None,
-        last_success=cc_pair.last_successful_index_time,
+        last_success=last_successful_index_time,
         docs_indexed=document_cnt,
         latest_index_attempt_docs_indexed=(
             latest_index_attempt.total_docs_indexed if latest_index_attempt else None

backend/onyx/server/documents/models.py

@@ -330,6 +330,7 @@ class CCPairFullInfo(BaseModel):
         num_docs_indexed: int,  # not ideal, but this must be computed separately
         is_editable_for_current_user: bool,
         indexing: bool,
+        last_successful_index_time: datetime | None = None,
         last_permission_sync_attempt_status: PermissionSyncStatus | None = None,
         permission_syncing: bool = False,
         last_permission_sync_attempt_finished: datetime | None = None,
@@ -382,9 +383,7 @@ class CCPairFullInfo(BaseModel):
             creator_email=(
                 cc_pair_model.creator.email if cc_pair_model.creator else None
             ),
-            last_indexed=(
-                last_index_attempt.time_started if last_index_attempt else None
-            ),
+            last_indexed=last_successful_index_time,
             last_pruned=cc_pair_model.last_pruned,
             last_full_permission_sync=cls._get_last_full_permission_sync(cc_pair_model),
             overall_indexing_speed=overall_indexing_speed,

backend/onyx/server/features/build/api/user_library.py

@@ -40,6 +40,8 @@ from sqlalchemy.orm import Session
 
 from onyx.auth.users import current_user
 from onyx.background.celery.versioned_apps.client import app as celery_app
+from onyx.configs.app_configs import MAX_EMBEDDED_IMAGES_PER_FILE
+from onyx.configs.app_configs import MAX_EMBEDDED_IMAGES_PER_UPLOAD
 from onyx.configs.constants import DocumentSource
 from onyx.configs.constants import OnyxCeleryQueues
 from onyx.configs.constants import OnyxCeleryTask
@@ -50,6 +52,9 @@ from onyx.db.engine.sql_engine import get_session
 from onyx.db.enums import ConnectorCredentialPairStatus
 from onyx.db.models import User
 from onyx.document_index.interfaces import DocumentMetadata
+from onyx.error_handling.error_codes import OnyxErrorCode
+from onyx.error_handling.exceptions import OnyxError
+from onyx.file_processing.extract_file_text import count_pdf_embedded_images
 from onyx.server.features.build.configs import USER_LIBRARY_MAX_FILE_SIZE_BYTES
 from onyx.server.features.build.configs import USER_LIBRARY_MAX_FILES_PER_UPLOAD
 from onyx.server.features.build.configs import USER_LIBRARY_MAX_TOTAL_SIZE_BYTES
@@ -127,6 +132,49 @@ class DeleteFileResponse(BaseModel):
 # =============================================================================
 
 
+def _looks_like_pdf(filename: str, content_type: str | None) -> bool:
+    """True if either the filename or the content-type indicates a PDF.
+
+    Client-supplied ``content_type`` can be spoofed (e.g. a PDF uploaded with
+    ``Content-Type: application/octet-stream``), so we also fall back to
+    extension-based detection via ``mimetypes.guess_type`` on the filename.
+    """
+    if content_type == "application/pdf":
+        return True
+    guessed, _ = mimetypes.guess_type(filename)
+    return guessed == "application/pdf"
+
+
+def _check_pdf_image_caps(
+    filename: str, content: bytes, content_type: str | None, batch_total: int
+) -> int:
+    """Enforce per-file and per-batch embedded-image caps for PDFs.
+
+    Returns the number of embedded images in this file (0 for non-PDFs) so
+    callers can update their running batch total. Raises OnyxError(INVALID_INPUT)
+    if either cap is exceeded.
+    """
+    if not _looks_like_pdf(filename, content_type):
+        return 0
+    file_cap = MAX_EMBEDDED_IMAGES_PER_FILE
+    batch_cap = MAX_EMBEDDED_IMAGES_PER_UPLOAD
+    # Short-circuit at the larger cap so we get a useful count for both checks.
+    count = count_pdf_embedded_images(BytesIO(content), max(file_cap, batch_cap))
+    if count > file_cap:
+        raise OnyxError(
+            OnyxErrorCode.INVALID_INPUT,
+            f"PDF '{filename}' contains too many embedded images "
+            f"(more than {file_cap}). Try splitting the document into smaller files.",
+        )
+    if batch_total + count > batch_cap:
+        raise OnyxError(
+            OnyxErrorCode.INVALID_INPUT,
+            f"Upload would exceed the {batch_cap}-image limit across all "
+            f"files in this batch. Try uploading fewer image-heavy files at once.",
+        )
+    return count
+
+
 def _sanitize_path(path: str) -> str:
     """Sanitize a file path, removing traversal attempts and normalizing.
 
@@ -356,6 +404,7 @@ async def upload_files(
 
     uploaded_entries: list[LibraryEntryResponse] = []
     total_size = 0
+    batch_image_total = 0
     now = datetime.now(timezone.utc)
 
     # Sanitize the base path
@@ -375,6 +424,14 @@ async def upload_files(
                 detail=f"File '{file.filename}' exceeds maximum size of {USER_LIBRARY_MAX_FILE_SIZE_BYTES // (1024*1024)}MB",
             )
 
+        # Reject PDFs with an unreasonable per-file or per-batch image count
+        batch_image_total += _check_pdf_image_caps(
+            filename=file.filename or "unnamed",
+            content=content,
+            content_type=file.content_type,
+            batch_total=batch_image_total,
+        )
+
         # Validate cumulative storage (existing + this upload batch)
         total_size += file_size
         if existing_usage + total_size > USER_LIBRARY_MAX_TOTAL_SIZE_BYTES:
@@ -473,6 +530,7 @@ async def upload_zip(
 
     uploaded_entries: list[LibraryEntryResponse] = []
     total_size = 0
+    batch_image_total = 0
 
     # Extract zip contents into a subfolder named after the zip file
     zip_name = api_sanitize_filename(file.filename or "upload")
@@ -511,6 +569,36 @@ async def upload_zip(
                     logger.warning(f"Skipping '{zip_info.filename}' - exceeds max size")
                     continue
 
+                # Skip PDFs that would trip the per-file or per-batch image
+                # cap (would OOM the user-file-processing worker). Matches
+                # /upload behavior but uses skip-and-warn to stay consistent
+                # with the zip path's handling of oversized files.
+                zip_file_name = zip_info.filename.split("/")[-1]
+                zip_content_type, _ = mimetypes.guess_type(zip_file_name)
+                if zip_content_type == "application/pdf":
+                    image_count = count_pdf_embedded_images(
+                        BytesIO(file_content),
+                        max(
+                            MAX_EMBEDDED_IMAGES_PER_FILE,
+                            MAX_EMBEDDED_IMAGES_PER_UPLOAD,
+                        ),
+                    )
+                    if image_count > MAX_EMBEDDED_IMAGES_PER_FILE:
+                        logger.warning(
+                            "Skipping '%s' - exceeds %d per-file embedded-image cap",
+                            zip_info.filename,
+                            MAX_EMBEDDED_IMAGES_PER_FILE,
+                        )
+                        continue
+                    if batch_image_total + image_count > MAX_EMBEDDED_IMAGES_PER_UPLOAD:
+                        logger.warning(
+                            "Skipping '%s' - would exceed %d per-batch embedded-image cap",
+                            zip_info.filename,
+                            MAX_EMBEDDED_IMAGES_PER_UPLOAD,
+                        )
+                        continue
+                    batch_image_total += image_count
+
                 total_size += file_size
 
                 # Validate cumulative storage

backend/onyx/server/features/projects/projects_file_utils.py

@@ -10,7 +10,10 @@ from pydantic import Field
 from sqlalchemy.orm import Session
 
 from onyx.configs.app_configs import FILE_TOKEN_COUNT_THRESHOLD
+from onyx.configs.app_configs import MAX_EMBEDDED_IMAGES_PER_FILE
+from onyx.configs.app_configs import MAX_EMBEDDED_IMAGES_PER_UPLOAD
 from onyx.db.llm import fetch_default_llm_model
+from onyx.file_processing.extract_file_text import count_pdf_embedded_images
 from onyx.file_processing.extract_file_text import extract_file_text
 from onyx.file_processing.extract_file_text import get_file_ext
 from onyx.file_processing.file_types import OnyxFileExtensions
@@ -156,6 +159,11 @@ def categorize_uploaded_files(
         except RuntimeError as e:
             logger.warning(f"Failed to get current tenant ID: {str(e)}")
 
+    # Running total of embedded images across PDFs in this batch. Once the
+    # aggregate cap is reached, subsequent PDFs in the same upload are
+    # rejected even if they'd individually fit under MAX_EMBEDDED_IMAGES_PER_FILE.
+    batch_image_total = 0
+
     for upload in files:
         try:
             filename = get_safe_filename(upload)
@@ -204,6 +212,47 @@ def categorize_uploaded_files(
                     )
                     continue
 
+                # Reject PDFs with an unreasonable number of embedded images
+                # (either per-file or accumulated across this upload batch).
+                # A PDF with thousands of embedded images can OOM the
+                # user-file-processing celery worker because every image is
+                # decoded with PIL and then sent to the vision LLM.
+                if extension == ".pdf":
+                    file_cap = MAX_EMBEDDED_IMAGES_PER_FILE
+                    batch_cap = MAX_EMBEDDED_IMAGES_PER_UPLOAD
+                    # Use the larger of the two caps as the short-circuit
+                    # threshold so we get a useful count for both checks.
+                    # count_pdf_embedded_images restores the stream position.
+                    count = count_pdf_embedded_images(
+                        upload.file, max(file_cap, batch_cap)
+                    )
+                    if count > file_cap:
+                        results.rejected.append(
+                            RejectedFile(
+                                filename=filename,
+                                reason=(
+                                    f"PDF contains too many embedded images "
+                                    f"(more than {file_cap}). Try splitting "
+                                    f"the document into smaller files."
+                                ),
+                            )
+                        )
+                        continue
+                    if batch_image_total + count > batch_cap:
+                        results.rejected.append(
+                            RejectedFile(
+                                filename=filename,
+                                reason=(
+                                    f"Upload would exceed the "
+                                    f"{batch_cap}-image limit across all "
+                                    f"files in this batch. Try uploading "
+                                    f"fewer image-heavy files at once."
+                                ),
+                            )
+                        )
+                        continue
+                    batch_image_total += count
+
                 text_content = extract_file_text(
                     file=upload.file,
                     file_name=filename,

backend/onyx/server/manage/llm/api.py

@@ -81,6 +81,7 @@ from onyx.server.manage.llm.models import VisionProviderResponse
 from onyx.server.manage.llm.utils import generate_bedrock_display_name
 from onyx.server.manage.llm.utils import generate_ollama_display_name
 from onyx.server.manage.llm.utils import infer_vision_support
+from onyx.server.manage.llm.utils import is_embedding_model
 from onyx.server.manage.llm.utils import is_reasoning_model
 from onyx.server.manage.llm.utils import is_valid_bedrock_model
 from onyx.server.manage.llm.utils import ModelMetadata
@@ -102,6 +103,43 @@ def _mask_string(value: str) -> str:
     return value[:4] + "****" + value[-4:]
 
 
+def _resolve_api_key(
+    api_key: str | None,
+    provider_name: str | None,
+    api_base: str | None,
+    db_session: Session,
+) -> str | None:
+    """Return the real API key for model-fetch endpoints.
+
+    When editing an existing provider the form value is masked (e.g.
+    ``sk-a****b1c2``).  If *provider_name* is supplied we can look up
+    the unmasked key from the database so the external request succeeds.
+
+    The stored key is only returned when the request's *api_base*
+    matches the value stored in the database.
+    """
+    if not provider_name:
+        return api_key
+
+    existing_provider = fetch_existing_llm_provider(
+        name=provider_name, db_session=db_session
+    )
+    if existing_provider and existing_provider.api_key:
+        # Normalise both URLs before comparing so trailing-slash
+        # differences don't cause a false mismatch.
+        stored_base = (existing_provider.api_base or "").strip().rstrip("/")
+        request_base = (api_base or "").strip().rstrip("/")
+        if stored_base != request_base:
+            return api_key
+
+        stored_key = existing_provider.api_key.get_value(apply_mask=False)
+        # Only resolve when the incoming value is the masked form of the
+        # stored key — i.e. the user hasn't typed a new key.
+        if api_key and api_key == _mask_string(stored_key):
+            return stored_key
+    return api_key
+
+
 def _sync_fetched_models(
     db_session: Session,
     provider_name: str,
@@ -1144,16 +1182,17 @@ def get_ollama_available_models(
     return sorted_results
 
 
-def _get_openrouter_models_response(api_base: str, api_key: str) -> dict:
+def _get_openrouter_models_response(api_base: str, api_key: str | None) -> dict:
     """Perform GET to OpenRouter /models and return parsed JSON."""
     cleaned_api_base = api_base.strip().rstrip("/")
     url = f"{cleaned_api_base}/models"
-    headers = {
-        "Authorization": f"Bearer {api_key}",
+    headers: dict[str, str] = {
         # Optional headers recommended by OpenRouter for attribution
         "HTTP-Referer": "https://onyx.app",
         "X-Title": "Onyx",
     }
+    if api_key:
+        headers["Authorization"] = f"Bearer {api_key}"
     try:
         response = httpx.get(url, headers=headers, timeout=10.0)
         response.raise_for_status()
@@ -1176,8 +1215,12 @@ def get_openrouter_available_models(
     Parses id, name (display), context_length, and architecture.input_modalities.
     """
 
+    api_key = _resolve_api_key(
+        request.api_key, request.provider_name, request.api_base, db_session
+    )
+
     response_json = _get_openrouter_models_response(
-        api_base=request.api_base, api_key=request.api_key
+        api_base=request.api_base, api_key=api_key
     )
 
     data = response_json.get("data", [])
@@ -1270,13 +1313,23 @@ def get_lm_studio_available_models(
 
     # If provider_name is given and the api_key hasn't been changed by the user,
     # fall back to the stored API key from the database (the form value is masked).
+    # Only do so when the api_base matches what is stored.
     api_key = request.api_key
     if request.provider_name and not request.api_key_changed:
         existing_provider = fetch_existing_llm_provider(
             name=request.provider_name, db_session=db_session
         )
         if existing_provider and existing_provider.custom_config:
-            api_key = existing_provider.custom_config.get(LM_STUDIO_API_KEY_CONFIG_KEY)
+            stored_base = (
+                (existing_provider.api_base or "")
+                .strip()
+                .rstrip("/")
+                .removesuffix("/v1")
+            )
+            if stored_base == cleaned_api_base:
+                api_key = existing_provider.custom_config.get(
+                    LM_STUDIO_API_KEY_CONFIG_KEY
+                )
 
     url = f"{cleaned_api_base}/api/v1/models"
     headers: dict[str, str] = {}
@@ -1360,8 +1413,12 @@ def get_litellm_available_models(
     db_session: Session = Depends(get_session),
 ) -> list[LitellmFinalModelResponse]:
     """Fetch available models from Litellm proxy /v1/models endpoint."""
+    api_key = _resolve_api_key(
+        request.api_key, request.provider_name, request.api_base, db_session
+    )
+
     response_json = _get_litellm_models_response(
-        api_key=request.api_key, api_base=request.api_base
+        api_key=api_key, api_base=request.api_base
     )
 
     models = response_json.get("data", [])
@@ -1376,6 +1433,10 @@ def get_litellm_available_models(
         try:
             model_details = LitellmModelDetails.model_validate(model)
 
+            # Skip embedding models
+            if is_embedding_model(model_details.id):
+                continue
+
             results.append(
                 LitellmFinalModelResponse(
                     provider_name=model_details.owned_by,
@@ -1414,16 +1475,17 @@ def get_litellm_available_models(
     return sorted_results
 
 
-def _get_litellm_models_response(api_key: str, api_base: str) -> dict:
+def _get_litellm_models_response(api_key: str | None, api_base: str) -> dict:
     """Perform GET to Litellm proxy /api/v1/models and return parsed JSON."""
     cleaned_api_base = api_base.strip().rstrip("/")
     url = f"{cleaned_api_base}/v1/models"
 
     headers = {
-        "Authorization": f"Bearer {api_key}",
         "HTTP-Referer": "https://onyx.app",
         "X-Title": "Onyx",
     }
+    if api_key:
+        headers["Authorization"] = f"Bearer {api_key}"
 
     try:
         response = httpx.get(url, headers=headers, timeout=10.0)

backend/onyx/server/manage/llm/utils.py

@@ -366,3 +366,18 @@ def extract_vendor_from_model_name(model_name: str, provider: str) -> str | None
         return None
 
     return None
+
+
+def is_embedding_model(model_name: str) -> bool:
+    """Checks for if a model is an embedding model"""
+    from litellm import get_model_info
+
+    try:
+        # get_model_info raises on unknown models
+        # default to False
+        model_info = get_model_info(model_name)
+    except Exception:
+        return False
+    is_embedding_mode = model_info.get("mode") == "embedding"
+
+    return is_embedding_mode

backend/onyx/tools/tool_implementations/python/python_tool.py

@@ -1,3 +1,4 @@
+import hashlib
 import mimetypes
 from io import BytesIO
 from typing import Any
@@ -85,6 +86,14 @@ class PythonTool(Tool[PythonToolOverrideKwargs]):
     def __init__(self, tool_id: int, emitter: Emitter) -> None:
         super().__init__(emitter=emitter)
         self._id = tool_id
+        # Cache of (filename, content_hash) -> ci_file_id to avoid re-uploading
+        # the same file on every tool call iteration within the same agent session.
+        # Filename is included in the key so two files with identical bytes but
+        # different names each get their own upload slot.
+        # TTL assumption: code-interpreter file TTLs (typically hours) greatly
+        # exceed the lifetime of a single agent session (at most MAX_LLM_CYCLES
+        # iterations, typically a few minutes), so stale-ID eviction is not needed.
+        self._uploaded_file_cache: dict[tuple[str, str], str] = {}
 
     @property
     def id(self) -> int:
@@ -184,8 +193,13 @@ class PythonTool(Tool[PythonToolOverrideKwargs]):
             for ind, chat_file in enumerate(chat_files):
                 file_name = chat_file.filename or f"file_{ind}"
                 try:
-                    # Upload to Code Interpreter
-                    ci_file_id = client.upload_file(chat_file.content, file_name)
+                    content_hash = hashlib.sha256(chat_file.content).hexdigest()
+                    cache_key = (file_name, content_hash)
+                    ci_file_id = self._uploaded_file_cache.get(cache_key)
+                    if ci_file_id is None:
+                        # Upload to Code Interpreter
+                        ci_file_id = client.upload_file(chat_file.content, file_name)
+                        self._uploaded_file_cache[cache_key] = ci_file_id
 
                     # Stage for execution
                     files_to_stage.append({"path": file_name, "file_id": ci_file_id})
@@ -303,15 +317,10 @@ class PythonTool(Tool[PythonToolOverrideKwargs]):
                             f"file {ci_file_id}: {e}"
                         )
 
-                # Cleanup staged input files
-                for file_mapping in files_to_stage:
-                    try:
-                        client.delete_file(file_mapping["file_id"])
-                    except Exception as e:
-                        logger.error(
-                            f"Failed to delete Code Interpreter staged "
-                            f"file {file_mapping['file_id']}: {e}"
-                        )
+                # Note: staged input files are intentionally not deleted here because
+                # _uploaded_file_cache reuses their file_ids across iterations. They are
+                # orphaned when the session ends, but the code interpreter cleans up
+                # stale files on its own TTL.
 
                 # Emit file_ids once files are processed
                 if generated_file_ids:

backend/requirements/default.txt

@@ -229,7 +229,9 @@ distro==1.9.0
 dnspython==2.8.0
     # via email-validator
 docstring-parser==0.17.0
-    # via cyclopts
+    # via
+    #   cyclopts
+    #   google-cloud-aiplatform
 docutils==0.22.3
     # via rich-rst
 dropbox==12.0.2
@@ -294,7 +296,13 @@ gitdb==4.0.12
 gitpython==3.1.45
     # via braintrust
 google-api-core==2.28.1
-    # via google-api-python-client
+    # via
+    #   google-api-python-client
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
+    #   google-cloud-core
+    #   google-cloud-resource-manager
+    #   google-cloud-storage
 google-api-python-client==2.86.0
     # via onyx
 google-auth==2.48.0
@@ -303,6 +311,11 @@ google-auth==2.48.0
     #   google-api-python-client
     #   google-auth-httplib2
     #   google-auth-oauthlib
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
+    #   google-cloud-core
+    #   google-cloud-resource-manager
+    #   google-cloud-storage
     #   google-genai
     #   kubernetes
 google-auth-httplib2==0.1.0
@@ -311,16 +324,51 @@ google-auth-httplib2==0.1.0
     #   onyx
 google-auth-oauthlib==1.0.0
     # via onyx
+google-cloud-aiplatform==1.133.0
+    # via litellm
+google-cloud-bigquery==3.41.0
+    # via google-cloud-aiplatform
+google-cloud-core==2.5.1
+    # via
+    #   google-cloud-bigquery
+    #   google-cloud-storage
+google-cloud-resource-manager==1.17.0
+    # via google-cloud-aiplatform
+google-cloud-storage==3.10.1
+    # via google-cloud-aiplatform
+google-crc32c==1.8.0
+    # via
+    #   google-cloud-storage
+    #   google-resumable-media
 google-genai==1.52.0
-    # via onyx
+    # via
+    #   google-cloud-aiplatform
+    #   onyx
+google-resumable-media==2.8.2
+    # via
+    #   google-cloud-bigquery
+    #   google-cloud-storage
 googleapis-common-protos==1.72.0
     # via
     #   google-api-core
+    #   grpc-google-iam-v1
+    #   grpcio-status
     #   opentelemetry-exporter-otlp-proto-http
 greenlet==3.2.4
     # via
     #   playwright
     #   sqlalchemy
+grpc-google-iam-v1==0.14.4
+    # via google-cloud-resource-manager
+grpcio==1.80.0
+    # via
+    #   google-api-core
+    #   google-cloud-resource-manager
+    #   googleapis-common-protos
+    #   grpc-google-iam-v1
+    #   grpcio-status
+grpcio-status==1.80.0
+    # via google-api-core
 h11==0.16.0
     # via
     #   httpcore
@@ -621,6 +669,8 @@ packaging==24.2
     #   dask
     #   distributed
     #   fastmcp
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
     #   huggingface-hub
     #   jira
     #   kombu
@@ -670,12 +720,19 @@ propcache==0.4.1
     #   aiohttp
     #   yarl
 proto-plus==1.26.1
-    # via google-api-core
+    # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-resource-manager
 protobuf==6.33.5
     # via
     #   ddtrace
     #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-resource-manager
     #   googleapis-common-protos
+    #   grpc-google-iam-v1
+    #   grpcio-status
     #   onnxruntime
     #   opentelemetry-proto
     #   proto-plus
@@ -713,6 +770,7 @@ pydantic==2.11.7
     #   exa-py
     #   fastapi
     #   fastmcp
+    #   google-cloud-aiplatform
     #   google-genai
     #   langchain-core
     #   langfuse
@@ -776,6 +834,7 @@ python-dateutil==2.8.2
     #   botocore
     #   celery
     #   dateparser
+    #   google-cloud-bigquery
     #   htmldate
     #   hubspot-api-client
     #   kubernetes
@@ -867,6 +926,8 @@ requests==2.32.5
     #   dropbox
     #   exa-py
     #   google-api-core
+    #   google-cloud-bigquery
+    #   google-cloud-storage
     #   google-genai
     #   hubspot-api-client
     #   huggingface-hub
@@ -1054,7 +1115,9 @@ typing-extensions==4.15.0
     #   exa-py
     #   exceptiongroup
     #   fastapi
+    #   google-cloud-aiplatform
     #   google-genai
+    #   grpcio
     #   huggingface-hub
     #   jira
     #   langchain-core

backend/requirements/dev.txt

@@ -115,6 +115,8 @@ distlib==0.4.0
     # via virtualenv
 distro==1.9.0
     # via openai
+docstring-parser==0.17.0
+    # via google-cloud-aiplatform
 durationpy==0.10
     # via kubernetes
 execnet==2.1.2
@@ -143,14 +145,65 @@ frozenlist==1.8.0
     #   aiosignal
 fsspec==2025.10.0
     # via huggingface-hub
+google-api-core==2.28.1
+    # via
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
+    #   google-cloud-core
+    #   google-cloud-resource-manager
+    #   google-cloud-storage
 google-auth==2.48.0
     # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
+    #   google-cloud-core
+    #   google-cloud-resource-manager
+    #   google-cloud-storage
     #   google-genai
     #   kubernetes
+google-cloud-aiplatform==1.133.0
+    # via litellm
+google-cloud-bigquery==3.41.0
+    # via google-cloud-aiplatform
+google-cloud-core==2.5.1
+    # via
+    #   google-cloud-bigquery
+    #   google-cloud-storage
+google-cloud-resource-manager==1.17.0
+    # via google-cloud-aiplatform
+google-cloud-storage==3.10.1
+    # via google-cloud-aiplatform
+google-crc32c==1.8.0
+    # via
+    #   google-cloud-storage
+    #   google-resumable-media
 google-genai==1.52.0
-    # via onyx
+    # via
+    #   google-cloud-aiplatform
+    #   onyx
+google-resumable-media==2.8.2
+    # via
+    #   google-cloud-bigquery
+    #   google-cloud-storage
+googleapis-common-protos==1.72.0
+    # via
+    #   google-api-core
+    #   grpc-google-iam-v1
+    #   grpcio-status
 greenlet==3.2.4 ; platform_machine == 'AMD64' or platform_machine == 'WIN32' or platform_machine == 'aarch64' or platform_machine == 'amd64' or platform_machine == 'ppc64le' or platform_machine == 'win32' or platform_machine == 'x86_64'
     # via sqlalchemy
+grpc-google-iam-v1==0.14.4
+    # via google-cloud-resource-manager
+grpcio==1.80.0
+    # via
+    #   google-api-core
+    #   google-cloud-resource-manager
+    #   googleapis-common-protos
+    #   grpc-google-iam-v1
+    #   grpcio-status
+grpcio-status==1.80.0
+    # via google-api-core
 h11==0.16.0
     # via
     #   httpcore
@@ -276,6 +329,8 @@ openapi-generator-cli==7.17.0
 packaging==24.2
     # via
     #   black
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
     #   hatchling
     #   huggingface-hub
     #   ipykernel
@@ -318,6 +373,20 @@ propcache==0.4.1
     # via
     #   aiohttp
     #   yarl
+proto-plus==1.26.1
+    # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-resource-manager
+protobuf==6.33.5
+    # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-resource-manager
+    #   googleapis-common-protos
+    #   grpc-google-iam-v1
+    #   grpcio-status
+    #   proto-plus
 psutil==7.1.3
     # via ipykernel
 ptyprocess==0.7.0 ; sys_platform != 'emscripten' and sys_platform != 'win32'
@@ -339,6 +408,7 @@ pydantic==2.11.7
     #   agent-client-protocol
     #   cohere
     #   fastapi
+    #   google-cloud-aiplatform
     #   google-genai
     #   litellm
     #   mcp
@@ -379,6 +449,7 @@ python-dateutil==2.8.2
     # via
     #   aiobotocore
     #   botocore
+    #   google-cloud-bigquery
     #   jupyter-client
     #   kubernetes
     #   matplotlib
@@ -413,6 +484,9 @@ reorder-python-imports-black==3.14.0
 requests==2.32.5
     # via
     #   cohere
+    #   google-api-core
+    #   google-cloud-bigquery
+    #   google-cloud-storage
     #   google-genai
     #   huggingface-hub
     #   kubernetes
@@ -525,7 +599,9 @@ typing-extensions==4.15.0
     #   celery-types
     #   cohere
     #   fastapi
+    #   google-cloud-aiplatform
     #   google-genai
+    #   grpcio
     #   huggingface-hub
     #   ipython
     #   mcp

backend/requirements/ee.txt

@@ -86,6 +86,8 @@ discord-py==2.4.0
     # via onyx
 distro==1.9.0
     # via openai
+docstring-parser==0.17.0
+    # via google-cloud-aiplatform
 durationpy==0.10
     # via kubernetes
 fastapi==0.133.1
@@ -102,12 +104,63 @@ frozenlist==1.8.0
     #   aiosignal
 fsspec==2025.10.0
     # via huggingface-hub
+google-api-core==2.28.1
+    # via
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
+    #   google-cloud-core
+    #   google-cloud-resource-manager
+    #   google-cloud-storage
 google-auth==2.48.0
     # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
+    #   google-cloud-core
+    #   google-cloud-resource-manager
+    #   google-cloud-storage
     #   google-genai
     #   kubernetes
+google-cloud-aiplatform==1.133.0
+    # via litellm
+google-cloud-bigquery==3.41.0
+    # via google-cloud-aiplatform
+google-cloud-core==2.5.1
+    # via
+    #   google-cloud-bigquery
+    #   google-cloud-storage
+google-cloud-resource-manager==1.17.0
+    # via google-cloud-aiplatform
+google-cloud-storage==3.10.1
+    # via google-cloud-aiplatform
+google-crc32c==1.8.0
+    # via
+    #   google-cloud-storage
+    #   google-resumable-media
 google-genai==1.52.0
-    # via onyx
+    # via
+    #   google-cloud-aiplatform
+    #   onyx
+google-resumable-media==2.8.2
+    # via
+    #   google-cloud-bigquery
+    #   google-cloud-storage
+googleapis-common-protos==1.72.0
+    # via
+    #   google-api-core
+    #   grpc-google-iam-v1
+    #   grpcio-status
+grpc-google-iam-v1==0.14.4
+    # via google-cloud-resource-manager
+grpcio==1.80.0
+    # via
+    #   google-api-core
+    #   google-cloud-resource-manager
+    #   googleapis-common-protos
+    #   grpc-google-iam-v1
+    #   grpcio-status
+grpcio-status==1.80.0
+    # via google-api-core
 h11==0.16.0
     # via
     #   httpcore
@@ -178,7 +231,10 @@ openai==2.14.0
     #   litellm
     #   onyx
 packaging==24.2
-    # via huggingface-hub
+    # via
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
+    #   huggingface-hub
 parameterized==0.9.0
     # via cohere
 posthog==3.7.4
@@ -193,6 +249,20 @@ propcache==0.4.1
     # via
     #   aiohttp
     #   yarl
+proto-plus==1.26.1
+    # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-resource-manager
+protobuf==6.33.5
+    # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-resource-manager
+    #   googleapis-common-protos
+    #   grpc-google-iam-v1
+    #   grpcio-status
+    #   proto-plus
 py==1.11.0
     # via retry
 pyasn1==0.6.2
@@ -208,6 +278,7 @@ pydantic==2.11.7
     #   agent-client-protocol
     #   cohere
     #   fastapi
+    #   google-cloud-aiplatform
     #   google-genai
     #   litellm
     #   mcp
@@ -224,6 +295,7 @@ python-dateutil==2.8.2
     # via
     #   aiobotocore
     #   botocore
+    #   google-cloud-bigquery
     #   kubernetes
     #   posthog
 python-dotenv==1.1.1
@@ -247,6 +319,9 @@ regex==2025.11.3
 requests==2.32.5
     # via
     #   cohere
+    #   google-api-core
+    #   google-cloud-bigquery
+    #   google-cloud-storage
     #   google-genai
     #   huggingface-hub
     #   kubernetes
@@ -306,7 +381,9 @@ typing-extensions==4.15.0
     #   anyio
     #   cohere
     #   fastapi
+    #   google-cloud-aiplatform
     #   google-genai
+    #   grpcio
     #   huggingface-hub
     #   mcp
     #   openai

backend/requirements/model_server.txt

@@ -102,6 +102,8 @@ discord-py==2.4.0
     # via onyx
 distro==1.9.0
     # via openai
+docstring-parser==0.17.0
+    # via google-cloud-aiplatform
 durationpy==0.10
     # via kubernetes
 einops==0.8.1
@@ -127,12 +129,63 @@ fsspec==2025.10.0
     # via
     #   huggingface-hub
     #   torch
+google-api-core==2.28.1
+    # via
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
+    #   google-cloud-core
+    #   google-cloud-resource-manager
+    #   google-cloud-storage
 google-auth==2.48.0
     # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
+    #   google-cloud-core
+    #   google-cloud-resource-manager
+    #   google-cloud-storage
     #   google-genai
     #   kubernetes
+google-cloud-aiplatform==1.133.0
+    # via litellm
+google-cloud-bigquery==3.41.0
+    # via google-cloud-aiplatform
+google-cloud-core==2.5.1
+    # via
+    #   google-cloud-bigquery
+    #   google-cloud-storage
+google-cloud-resource-manager==1.17.0
+    # via google-cloud-aiplatform
+google-cloud-storage==3.10.1
+    # via google-cloud-aiplatform
+google-crc32c==1.8.0
+    # via
+    #   google-cloud-storage
+    #   google-resumable-media
 google-genai==1.52.0
-    # via onyx
+    # via
+    #   google-cloud-aiplatform
+    #   onyx
+google-resumable-media==2.8.2
+    # via
+    #   google-cloud-bigquery
+    #   google-cloud-storage
+googleapis-common-protos==1.72.0
+    # via
+    #   google-api-core
+    #   grpc-google-iam-v1
+    #   grpcio-status
+grpc-google-iam-v1==0.14.4
+    # via google-cloud-resource-manager
+grpcio==1.80.0
+    # via
+    #   google-api-core
+    #   google-cloud-resource-manager
+    #   googleapis-common-protos
+    #   grpc-google-iam-v1
+    #   grpcio-status
+grpcio-status==1.80.0
+    # via google-api-core
 h11==0.16.0
     # via
     #   httpcore
@@ -262,6 +315,8 @@ openai==2.14.0
 packaging==24.2
     # via
     #   accelerate
+    #   google-cloud-aiplatform
+    #   google-cloud-bigquery
     #   huggingface-hub
     #   kombu
     #   transformers
@@ -281,6 +336,20 @@ propcache==0.4.1
     # via
     #   aiohttp
     #   yarl
+proto-plus==1.26.1
+    # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-resource-manager
+protobuf==6.33.5
+    # via
+    #   google-api-core
+    #   google-cloud-aiplatform
+    #   google-cloud-resource-manager
+    #   googleapis-common-protos
+    #   grpc-google-iam-v1
+    #   grpcio-status
+    #   proto-plus
 psutil==7.1.3
     # via accelerate
 py==1.11.0
@@ -298,6 +367,7 @@ pydantic==2.11.7
     #   agent-client-protocol
     #   cohere
     #   fastapi
+    #   google-cloud-aiplatform
     #   google-genai
     #   litellm
     #   mcp
@@ -315,6 +385,7 @@ python-dateutil==2.8.2
     #   aiobotocore
     #   botocore
     #   celery
+    #   google-cloud-bigquery
     #   kubernetes
 python-dotenv==1.1.1
     # via
@@ -341,6 +412,9 @@ regex==2025.11.3
 requests==2.32.5
     # via
     #   cohere
+    #   google-api-core
+    #   google-cloud-bigquery
+    #   google-cloud-storage
     #   google-genai
     #   huggingface-hub
     #   kubernetes
@@ -433,7 +507,9 @@ typing-extensions==4.15.0
     #   anyio
     #   cohere
     #   fastapi
+    #   google-cloud-aiplatform
     #   google-genai
+    #   grpcio
     #   huggingface-hub
     #   mcp
     #   openai

backend/tests/README.md

@@ -45,6 +45,21 @@ npx playwright test <TEST_NAME>
 Shared fixtures live in `backend/tests/conftest.py`. Test subdirectories can define
 their own `conftest.py` for directory-scoped fixtures.
 
+## Running Tests Repeatedly (`pytest-repeat`)
+
+Use `pytest-repeat` to catch flaky tests by running them multiple times:
+
+```bash
+# Run a specific test 50 times
+pytest --count=50 backend/tests/unit/path/to/test.py::test_name
+
+# Stop on first failure with -x
+pytest --count=50 -x backend/tests/unit/path/to/test.py::test_name
+
+# Repeat an entire test file
+pytest --count=10 backend/tests/unit/path/to/test_file.py
+```
+
 ## Best Practices
 
 ### Use `enable_ee` fixture instead of inlining

backend/tests/external_dependency_unit/celery/test_user_file_delete_queue.py

@@ -0,0 +1,274 @@
+"""
+External dependency unit tests for user file delete queue protections.
+
+Verifies that the three mechanisms added to check_for_user_file_delete work
+correctly:
+
+1. Queue depth backpressure – when the broker queue exceeds
+   USER_FILE_DELETE_MAX_QUEUE_DEPTH, no new tasks are enqueued.
+
+2. Per-file Redis guard key – if the guard key for a file already exists in
+   Redis, that file is skipped even though it is still in DELETING status.
+
+3. Task expiry – every send_task call carries expires=
+   CELERY_USER_FILE_DELETE_TASK_EXPIRES so that stale queued tasks are
+   discarded by workers automatically.
+
+Also verifies that delete_user_file_impl clears the guard key the moment
+it is picked up by a worker.
+
+Uses real Redis (DB 0 via get_redis_client) and real PostgreSQL for UserFile
+rows.  The Celery app is provided as a MagicMock injected via a PropertyMock
+on the task class so no real broker is needed.
+"""
+
+from collections.abc import Generator
+from contextlib import contextmanager
+from typing import Any
+from unittest.mock import MagicMock
+from unittest.mock import patch
+from unittest.mock import PropertyMock
+from uuid import uuid4
+
+from sqlalchemy.orm import Session
+
+from onyx.background.celery.tasks.user_file_processing.tasks import (
+    _user_file_delete_lock_key,
+)
+from onyx.background.celery.tasks.user_file_processing.tasks import (
+    _user_file_delete_queued_key,
+)
+from onyx.background.celery.tasks.user_file_processing.tasks import (
+    check_for_user_file_delete,
+)
+from onyx.background.celery.tasks.user_file_processing.tasks import (
+    process_single_user_file_delete,
+)
+from onyx.configs.constants import CELERY_USER_FILE_DELETE_TASK_EXPIRES
+from onyx.configs.constants import OnyxCeleryQueues
+from onyx.configs.constants import OnyxCeleryTask
+from onyx.configs.constants import USER_FILE_DELETE_MAX_QUEUE_DEPTH
+from onyx.db.enums import UserFileStatus
+from onyx.db.models import UserFile
+from onyx.redis.redis_pool import get_redis_client
+from tests.external_dependency_unit.conftest import create_test_user
+from tests.external_dependency_unit.constants import TEST_TENANT_ID
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+_PATCH_QUEUE_LEN = (
+    "onyx.background.celery.tasks.user_file_processing.tasks.celery_get_queue_length"
+)
+
+
+def _create_deleting_user_file(db_session: Session, user_id: object) -> UserFile:
+    """Insert a UserFile in DELETING status and return it."""
+    uf = UserFile(
+        id=uuid4(),
+        user_id=user_id,
+        file_id=f"test_file_{uuid4().hex[:8]}",
+        name=f"test_{uuid4().hex[:8]}.txt",
+        file_type="text/plain",
+        status=UserFileStatus.DELETING,
+    )
+    db_session.add(uf)
+    db_session.commit()
+    db_session.refresh(uf)
+    return uf
+
+
+@contextmanager
+def _patch_task_app(task: Any, mock_app: MagicMock) -> Generator[None, None, None]:
+    """Patch the ``app`` property on *task*'s class so that ``self.app``
+    inside the task function returns *mock_app*.
+
+    With ``bind=True``, ``task.run`` is a bound method whose ``__self__`` is
+    the actual task instance.  We patch ``app`` on that instance's class
+    (a unique Celery-generated Task subclass) so the mock is scoped to this
+    task only.
+    """
+    task_instance = task.run.__self__
+    with patch.object(
+        type(task_instance), "app", new_callable=PropertyMock, return_value=mock_app
+    ):
+        yield
+
+
+# ---------------------------------------------------------------------------
+# Test classes
+# ---------------------------------------------------------------------------
+
+
+class TestDeleteQueueDepthBackpressure:
+    """Protection 1: skip all enqueuing when the broker queue is too deep."""
+
+    def test_no_tasks_enqueued_when_queue_over_limit(
+        self,
+        db_session: Session,
+        tenant_context: None,  # noqa: ARG002
+    ) -> None:
+        """When the queue depth exceeds the limit the beat cycle is skipped."""
+        user = create_test_user(db_session, "del_bp_user")
+        _create_deleting_user_file(db_session, user.id)
+
+        mock_app = MagicMock()
+
+        with (
+            _patch_task_app(check_for_user_file_delete, mock_app),
+            patch(_PATCH_QUEUE_LEN, return_value=USER_FILE_DELETE_MAX_QUEUE_DEPTH + 1),
+        ):
+            check_for_user_file_delete.run(tenant_id=TEST_TENANT_ID)
+
+        mock_app.send_task.assert_not_called()
+
+
+class TestDeletePerFileGuardKey:
+    """Protection 2: per-file Redis guard key prevents duplicate enqueue."""
+
+    def test_guarded_file_not_re_enqueued(
+        self,
+        db_session: Session,
+        tenant_context: None,  # noqa: ARG002
+    ) -> None:
+        """A file whose guard key is already set in Redis is skipped."""
+        user = create_test_user(db_session, "del_guard_user")
+        uf = _create_deleting_user_file(db_session, user.id)
+
+        redis_client = get_redis_client(tenant_id=TEST_TENANT_ID)
+        guard_key = _user_file_delete_queued_key(uf.id)
+        redis_client.setex(guard_key, CELERY_USER_FILE_DELETE_TASK_EXPIRES, 1)
+
+        mock_app = MagicMock()
+
+        try:
+            with (
+                _patch_task_app(check_for_user_file_delete, mock_app),
+                patch(_PATCH_QUEUE_LEN, return_value=0),
+            ):
+                check_for_user_file_delete.run(tenant_id=TEST_TENANT_ID)
+
+            # send_task must not have been called with this specific file's ID
+            for call in mock_app.send_task.call_args_list:
+                kwargs = call.kwargs.get("kwargs", {})
+                assert kwargs.get("user_file_id") != str(
+                    uf.id
+                ), f"File {uf.id} should have been skipped because its guard key exists"
+        finally:
+            redis_client.delete(guard_key)
+
+    def test_guard_key_exists_in_redis_after_enqueue(
+        self,
+        db_session: Session,
+        tenant_context: None,  # noqa: ARG002
+    ) -> None:
+        """After a file is enqueued its guard key is present in Redis with a TTL."""
+        user = create_test_user(db_session, "del_guard_set_user")
+        uf = _create_deleting_user_file(db_session, user.id)
+
+        redis_client = get_redis_client(tenant_id=TEST_TENANT_ID)
+        guard_key = _user_file_delete_queued_key(uf.id)
+        redis_client.delete(guard_key)  # clean slate
+
+        mock_app = MagicMock()
+
+        try:
+            with (
+                _patch_task_app(check_for_user_file_delete, mock_app),
+                patch(_PATCH_QUEUE_LEN, return_value=0),
+            ):
+                check_for_user_file_delete.run(tenant_id=TEST_TENANT_ID)
+
+            assert redis_client.exists(
+                guard_key
+            ), "Guard key should be set in Redis after enqueue"
+            ttl = int(redis_client.ttl(guard_key))  # type: ignore[arg-type]
+            assert (
+                0 < ttl <= CELERY_USER_FILE_DELETE_TASK_EXPIRES
+            ), f"Guard key TTL {ttl}s is outside the expected range (0, {CELERY_USER_FILE_DELETE_TASK_EXPIRES}]"
+        finally:
+            redis_client.delete(guard_key)
+
+
+class TestDeleteTaskExpiry:
+    """Protection 3: every send_task call includes an expires value."""
+
+    def test_send_task_called_with_expires(
+        self,
+        db_session: Session,
+        tenant_context: None,  # noqa: ARG002
+    ) -> None:
+        """send_task is called with the correct queue, task name, and expires."""
+        user = create_test_user(db_session, "del_expires_user")
+        uf = _create_deleting_user_file(db_session, user.id)
+
+        redis_client = get_redis_client(tenant_id=TEST_TENANT_ID)
+        guard_key = _user_file_delete_queued_key(uf.id)
+        redis_client.delete(guard_key)
+
+        mock_app = MagicMock()
+
+        try:
+            with (
+                _patch_task_app(check_for_user_file_delete, mock_app),
+                patch(_PATCH_QUEUE_LEN, return_value=0),
+            ):
+                check_for_user_file_delete.run(tenant_id=TEST_TENANT_ID)
+
+            # At least one task should have been submitted (for our file)
+            assert (
+                mock_app.send_task.call_count >= 1
+            ), "Expected at least one task to be submitted"
+
+            # Every submitted task must carry expires
+            for call in mock_app.send_task.call_args_list:
+                assert call.args[0] == OnyxCeleryTask.DELETE_SINGLE_USER_FILE
+                assert call.kwargs.get("queue") == OnyxCeleryQueues.USER_FILE_DELETE
+                assert (
+                    call.kwargs.get("expires") == CELERY_USER_FILE_DELETE_TASK_EXPIRES
+                ), "Task must be submitted with the correct expires value to prevent stale task accumulation"
+        finally:
+            redis_client.delete(guard_key)
+
+
+class TestDeleteWorkerClearsGuardKey:
+    """process_single_user_file_delete removes the guard key when it picks up a task."""
+
+    def test_guard_key_deleted_on_pickup(
+        self,
+        tenant_context: None,  # noqa: ARG002
+    ) -> None:
+        """The guard key is deleted before the worker does any real work.
+
+        We simulate an already-locked file so delete_user_file_impl returns
+        early – but crucially, after the guard key deletion.
+        """
+        user_file_id = str(uuid4())
+
+        redis_client = get_redis_client(tenant_id=TEST_TENANT_ID)
+        guard_key = _user_file_delete_queued_key(user_file_id)
+
+        # Simulate the guard key set when the beat enqueued the task
+        redis_client.setex(guard_key, CELERY_USER_FILE_DELETE_TASK_EXPIRES, 1)
+        assert redis_client.exists(guard_key), "Guard key must exist before pickup"
+
+        # Hold the per-file delete lock so the worker exits early without
+        # touching the database or file store.
+        lock_key = _user_file_delete_lock_key(user_file_id)
+        delete_lock = redis_client.lock(lock_key, timeout=10)
+        acquired = delete_lock.acquire(blocking=False)
+        assert acquired, "Should be able to acquire the delete lock for this test"
+
+        try:
+            process_single_user_file_delete.run(
+                user_file_id=user_file_id,
+                tenant_id=TEST_TENANT_ID,
+            )
+        finally:
+            if delete_lock.owned():
+                delete_lock.release()
+
+        assert not redis_client.exists(
+            guard_key
+        ), "Guard key should be deleted when the worker picks up the task"

backend/tests/external_dependency_unit/tools/test_python_tool.py

@@ -1218,15 +1218,16 @@ def test_code_interpreter_receives_chat_files(
         finally:
             ci_mod.CodeInterpreterClient.__init__.__defaults__ = original_defaults
 
-    # Verify: file uploaded, code executed via streaming, staged file cleaned up
+    # Verify: file uploaded and code executed via streaming.
     assert len(mock_ci_server.get_requests(method="POST", path="/v1/files")) == 1
     assert (
         len(mock_ci_server.get_requests(method="POST", path="/v1/execute/stream")) == 1
     )
 
-    delete_requests = mock_ci_server.get_requests(method="DELETE")
-    assert len(delete_requests) == 1
-    assert delete_requests[0].path.startswith("/v1/files/")
+    # Staged input files are intentionally NOT deleted — PythonTool caches their
+    # file IDs across agent-loop iterations to avoid re-uploading on every call.
+    # The code interpreter cleans them up via its own TTL.
+    assert len(mock_ci_server.get_requests(method="DELETE")) == 0
 
     execute_body = mock_ci_server.get_requests(
         method="POST", path="/v1/execute/stream"

backend/tests/integration/tests/connector/test_last_indexed_time.py

@@ -0,0 +1,237 @@
+"""
+Integration tests for the "Last Indexed" time displayed on both the
+per-connector detail page and the all-connectors listing page.
+
+Expected behavior: "Last Indexed" = time_started of the most recent
+successful index attempt for the cc pair, regardless of pagination.
+
+Edge cases:
+1. First page of index attempts is entirely errors — last_indexed should
+   still reflect the older successful attempt beyond page 1.
+2. Credential swap — successful attempts, then failures after a
+   "credential change"; last_indexed should reflect the most recent
+   successful attempt.
+3. Mix of statuses — only the most recent successful attempt matters.
+4. COMPLETED_WITH_ERRORS counts as a success for last_indexed purposes.
+"""
+
+from datetime import datetime
+from datetime import timedelta
+from datetime import timezone
+
+from onyx.db.models import IndexingStatus
+from onyx.server.documents.models import CCPairFullInfo
+from onyx.server.documents.models import ConnectorIndexingStatusLite
+from tests.integration.common_utils.managers.cc_pair import CCPairManager
+from tests.integration.common_utils.managers.connector import ConnectorManager
+from tests.integration.common_utils.managers.credential import CredentialManager
+from tests.integration.common_utils.managers.index_attempt import IndexAttemptManager
+from tests.integration.common_utils.managers.user import UserManager
+from tests.integration.common_utils.test_models import DATestCCPair
+from tests.integration.common_utils.test_models import DATestUser
+
+
+def _wait_for_real_success(
+    cc_pair: DATestCCPair,
+    admin: DATestUser,
+) -> None:
+    """Wait for the initial index attempt to complete successfully."""
+    CCPairManager.wait_for_indexing_completion(
+        cc_pair,
+        after=datetime(2000, 1, 1, tzinfo=timezone.utc),
+        user_performing_action=admin,
+        timeout=120,
+    )
+
+
+def _get_detail(cc_pair_id: int, admin: DATestUser) -> CCPairFullInfo:
+    result = CCPairManager.get_single(cc_pair_id, admin)
+    assert result is not None
+    return result
+
+
+def _get_listing(cc_pair_id: int, admin: DATestUser) -> ConnectorIndexingStatusLite:
+    result = CCPairManager.get_indexing_status_by_id(cc_pair_id, admin)
+    assert result is not None
+    return result
+
+
+def test_last_indexed_first_page_all_errors(reset: None) -> None:  # noqa: ARG001
+    """When the first page of index attempts is entirely errors but an
+    older successful attempt exists, both the detail page and the listing
+    page should still show the time of that successful attempt.
+
+    The detail page UI uses page size 8. We insert 10 failed attempts
+    more recent than the initial success to push the success off page 1.
+    """
+    admin = UserManager.create(name="admin_first_page_errors")
+    cc_pair = CCPairManager.create_from_scratch(user_performing_action=admin)
+    _wait_for_real_success(cc_pair, admin)
+
+    # Baseline: last_success should be set from the initial successful run
+    listing_before = _get_listing(cc_pair.id, admin)
+    assert listing_before.last_success is not None
+
+    # 10 recent failures push the success off page 1
+    IndexAttemptManager.create_test_index_attempts(
+        num_attempts=10,
+        cc_pair_id=cc_pair.id,
+        status=IndexingStatus.FAILED,
+        error_msg="simulated failure",
+        base_time=datetime.now(tz=timezone.utc),
+    )
+
+    detail = _get_detail(cc_pair.id, admin)
+    listing = _get_listing(cc_pair.id, admin)
+
+    assert (
+        detail.last_indexed is not None
+    ), "Detail page last_indexed is None even though a successful attempt exists"
+    assert (
+        listing.last_success is not None
+    ), "Listing page last_success is None even though a successful attempt exists"
+
+    # Both surfaces must agree
+    assert detail.last_indexed == listing.last_success, (
+        f"Detail last_indexed={detail.last_indexed} != "
+        f"listing last_success={listing.last_success}"
+    )
+
+
+def test_last_indexed_credential_swap_scenario(reset: None) -> None:  # noqa: ARG001
+    """Perform an actual credential swap: create connector + cred1 (cc_pair_1),
+    wait for success, then associate a new cred2 with the same connector
+    (cc_pair_2), wait for that to succeed, and inject failures on cc_pair_2.
+
+    cc_pair_2's last_indexed must reflect cc_pair_2's own success, not
+    cc_pair_1's older one. Both the detail page and listing page must agree.
+    """
+    admin = UserManager.create(name="admin_cred_swap")
+
+    connector = ConnectorManager.create(user_performing_action=admin)
+    cred1 = CredentialManager.create(user_performing_action=admin)
+    cc_pair_1 = CCPairManager.create(
+        connector_id=connector.id,
+        credential_id=cred1.id,
+        user_performing_action=admin,
+    )
+    _wait_for_real_success(cc_pair_1, admin)
+
+    cred2 = CredentialManager.create(user_performing_action=admin, name="swapped-cred")
+    cc_pair_2 = CCPairManager.create(
+        connector_id=connector.id,
+        credential_id=cred2.id,
+        user_performing_action=admin,
+    )
+    _wait_for_real_success(cc_pair_2, admin)
+
+    listing_after_swap = _get_listing(cc_pair_2.id, admin)
+    assert listing_after_swap.last_success is not None
+
+    IndexAttemptManager.create_test_index_attempts(
+        num_attempts=10,
+        cc_pair_id=cc_pair_2.id,
+        status=IndexingStatus.FAILED,
+        error_msg="credential expired",
+        base_time=datetime.now(tz=timezone.utc),
+    )
+
+    detail = _get_detail(cc_pair_2.id, admin)
+    listing = _get_listing(cc_pair_2.id, admin)
+
+    assert detail.last_indexed is not None
+    assert listing.last_success is not None
+
+    assert detail.last_indexed == listing.last_success, (
+        f"Detail last_indexed={detail.last_indexed} != "
+        f"listing last_success={listing.last_success}"
+    )
+
+
+def test_last_indexed_mixed_statuses(reset: None) -> None:  # noqa: ARG001
+    """Mix of in_progress, failed, and successful attempts. Only the most
+    recent successful attempt's time matters."""
+    admin = UserManager.create(name="admin_mixed")
+    cc_pair = CCPairManager.create_from_scratch(user_performing_action=admin)
+    _wait_for_real_success(cc_pair, admin)
+
+    now = datetime.now(tz=timezone.utc)
+
+    # Success 5 hours ago
+    IndexAttemptManager.create_test_index_attempts(
+        num_attempts=1,
+        cc_pair_id=cc_pair.id,
+        status=IndexingStatus.SUCCESS,
+        base_time=now - timedelta(hours=5),
+    )
+
+    # Failures 3 hours ago
+    IndexAttemptManager.create_test_index_attempts(
+        num_attempts=3,
+        cc_pair_id=cc_pair.id,
+        status=IndexingStatus.FAILED,
+        error_msg="transient failure",
+        base_time=now - timedelta(hours=3),
+    )
+
+    # In-progress 1 hour ago
+    IndexAttemptManager.create_test_index_attempts(
+        num_attempts=1,
+        cc_pair_id=cc_pair.id,
+        status=IndexingStatus.IN_PROGRESS,
+        base_time=now - timedelta(hours=1),
+    )
+
+    detail = _get_detail(cc_pair.id, admin)
+    listing = _get_listing(cc_pair.id, admin)
+
+    assert detail.last_indexed is not None
+    assert listing.last_success is not None
+
+    assert detail.last_indexed == listing.last_success, (
+        f"Detail last_indexed={detail.last_indexed} != "
+        f"listing last_success={listing.last_success}"
+    )
+
+
+def test_last_indexed_completed_with_errors(reset: None) -> None:  # noqa: ARG001
+    """COMPLETED_WITH_ERRORS is treated as a successful attempt (matching
+    IndexingStatus.is_successful()). When it is the most recent "success"
+    and later attempts all failed, both surfaces should reflect its time."""
+    admin = UserManager.create(name="admin_completed_errors")
+    cc_pair = CCPairManager.create_from_scratch(user_performing_action=admin)
+    _wait_for_real_success(cc_pair, admin)
+
+    now = datetime.now(tz=timezone.utc)
+
+    # COMPLETED_WITH_ERRORS 2 hours ago
+    IndexAttemptManager.create_test_index_attempts(
+        num_attempts=1,
+        cc_pair_id=cc_pair.id,
+        status=IndexingStatus.COMPLETED_WITH_ERRORS,
+        base_time=now - timedelta(hours=2),
+    )
+
+    # 10 failures after — push everything else off page 1
+    IndexAttemptManager.create_test_index_attempts(
+        num_attempts=10,
+        cc_pair_id=cc_pair.id,
+        status=IndexingStatus.FAILED,
+        error_msg="post-partial failure",
+        base_time=now,
+    )
+
+    detail = _get_detail(cc_pair.id, admin)
+    listing = _get_listing(cc_pair.id, admin)
+
+    assert (
+        detail.last_indexed is not None
+    ), "COMPLETED_WITH_ERRORS should count as a success for last_indexed"
+    assert (
+        listing.last_success is not None
+    ), "COMPLETED_WITH_ERRORS should count as a success for last_success"
+
+    assert detail.last_indexed == listing.last_success, (
+        f"Detail last_indexed={detail.last_indexed} != "
+        f"listing last_success={listing.last_success}"
+    )

backend/tests/unit/onyx/db/test_projects_upload_task_expiry.py

@@ -0,0 +1,63 @@
+"""
+Unit test verifying that the upload API path sends tasks with expires=.
+
+The upload_files_to_user_files_with_indexing function must include expires=
+on every send_task call to prevent phantom task accumulation if the worker
+is down or slow.
+"""
+
+from unittest.mock import MagicMock
+from unittest.mock import patch
+from uuid import uuid4
+
+from onyx.configs.constants import CELERY_USER_FILE_PROCESSING_TASK_EXPIRES
+from onyx.configs.constants import OnyxCeleryQueues
+from onyx.configs.constants import OnyxCeleryTask
+from onyx.db.models import UserFile
+from onyx.db.projects import upload_files_to_user_files_with_indexing
+
+
+def _make_mock_user_file() -> MagicMock:
+    uf = MagicMock(spec=UserFile)
+    uf.id = str(uuid4())
+    return uf
+
+
+@patch("onyx.db.projects.get_current_tenant_id", return_value="test_tenant")
+@patch("onyx.db.projects.create_user_files")
+@patch(
+    "onyx.background.celery.versioned_apps.client.app",
+    new_callable=MagicMock,
+)
+def test_send_task_includes_expires(
+    mock_client_app: MagicMock,
+    mock_create: MagicMock,
+    mock_tenant: MagicMock,  # noqa: ARG001
+) -> None:
+    """Every send_task call from the upload path must include expires=."""
+    user_files = [_make_mock_user_file(), _make_mock_user_file()]
+    mock_create.return_value = MagicMock(
+        user_files=user_files,
+        rejected_files=[],
+        id_to_temp_id={},
+    )
+
+    mock_user = MagicMock()
+    mock_db_session = MagicMock()
+
+    upload_files_to_user_files_with_indexing(
+        files=[],
+        project_id=None,
+        user=mock_user,
+        temp_id_map=None,
+        db_session=mock_db_session,
+    )
+
+    assert mock_client_app.send_task.call_count == len(user_files)
+
+    for call in mock_client_app.send_task.call_args_list:
+        assert call.args[0] == OnyxCeleryTask.PROCESS_SINGLE_USER_FILE
+        assert call.kwargs.get("queue") == OnyxCeleryQueues.USER_FILE_PROCESSING
+        assert (
+            call.kwargs.get("expires") == CELERY_USER_FILE_PROCESSING_TASK_EXPIRES
+        ), "send_task must include expires= to prevent phantom task accumulation"

backend/tests/unit/onyx/file_processing/test_image_summarization_errors.py

@@ -0,0 +1,89 @@
+"""
+Unit tests for image summarization error handling.
+
+Verifies that:
+1. LLM errors produce actionable error messages (not base64 dumps)
+2. Unsupported MIME type logs include the magic bytes and size
+3. The ValueError raised on LLM failure preserves the original exception
+"""
+
+from unittest.mock import MagicMock
+from unittest.mock import patch
+
+import pytest
+
+from onyx.file_processing.image_summarization import _summarize_image
+from onyx.file_processing.image_summarization import summarize_image_with_error_handling
+
+
+class TestSummarizeImageErrorMessage:
+    """_summarize_image must not dump base64 image data into error messages."""
+
+    def test_error_message_contains_exception_type_not_base64(self) -> None:
+        """The ValueError should contain the original exception info, not message payloads."""
+        mock_llm = MagicMock()
+        mock_llm.invoke.side_effect = RuntimeError("Connection timeout")
+
+        # A fake base64-encoded image string (should NOT appear in the error)
+        fake_encoded = "data:image/png;base64,iVBORw0KGgoAAAANSUhEUg..."
+
+        with pytest.raises(ValueError, match="RuntimeError: Connection timeout"):
+            _summarize_image(fake_encoded, mock_llm, query="test")
+
+    def test_error_message_does_not_contain_base64(self) -> None:
+        """Ensure base64 data is never included in the error message."""
+        mock_llm = MagicMock()
+        mock_llm.invoke.side_effect = RuntimeError("API error")
+
+        fake_encoded = "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAUA"
+
+        with pytest.raises(ValueError) as exc_info:
+            _summarize_image(fake_encoded, mock_llm)
+
+        error_str = str(exc_info.value)
+        assert "base64" not in error_str
+        assert "iVBOR" not in error_str
+
+    def test_original_exception_is_chained(self) -> None:
+        """The ValueError should chain the original exception via __cause__."""
+        mock_llm = MagicMock()
+        original = RuntimeError("upstream failure")
+        mock_llm.invoke.side_effect = original
+
+        with pytest.raises(ValueError) as exc_info:
+            _summarize_image("data:image/png;base64,abc", mock_llm)
+
+        assert exc_info.value.__cause__ is original
+
+
+class TestUnsupportedMimeTypeLogging:
+    """summarize_image_with_error_handling should log useful info for unsupported formats."""
+
+    @patch(
+        "onyx.file_processing.image_summarization.summarize_image_pipeline",
+        side_effect=__import__(
+            "onyx.file_processing.image_summarization",
+            fromlist=["UnsupportedImageFormatError"],
+        ).UnsupportedImageFormatError("unsupported"),
+    )
+    def test_logs_magic_bytes_and_size(
+        self, mock_pipeline: MagicMock  # noqa: ARG002
+    ) -> None:
+        """The info log should include magic bytes hex and image size."""
+        mock_llm = MagicMock()
+        # TIFF magic bytes (not in the supported list)
+        image_data = b"\x49\x49\x2a\x00" + b"\x00" * 100
+
+        with patch("onyx.file_processing.image_summarization.logger") as mock_logger:
+            result = summarize_image_with_error_handling(
+                llm=mock_llm,
+                image_data=image_data,
+                context_name="test_image.tiff",
+            )
+
+        assert result is None
+        mock_logger.info.assert_called_once()
+        log_args = mock_logger.info.call_args
+        # Check the format string args contain magic bytes and size
+        assert "49492a00" in str(log_args)
+        assert "104" in str(log_args)  # 4 + 100 bytes

backend/tests/unit/onyx/file_processing/test_image_summarization_litellm_errors.py

@@ -0,0 +1,141 @@
+"""
+Unit tests verifying that LiteLLM error details are extracted and surfaced
+in image summarization error messages.
+
+When the LLM call fails, the error handler should include the status_code,
+llm_provider, and model from LiteLLM exceptions so operators can diagnose
+the root cause (rate limit, content filter, unsupported vision, etc.)
+without needing to dig through LiteLLM internals.
+"""
+
+from unittest.mock import MagicMock
+
+import pytest
+
+from onyx.file_processing.image_summarization import _summarize_image
+
+
+def _make_litellm_style_error(
+    *,
+    message: str = "API error",
+    status_code: int | None = None,
+    llm_provider: str | None = None,
+    model: str | None = None,
+) -> RuntimeError:
+    """Create an exception with LiteLLM-style attributes."""
+    exc = RuntimeError(message)
+    if status_code is not None:
+        exc.status_code = status_code  # type: ignore[attr-defined]
+    if llm_provider is not None:
+        exc.llm_provider = llm_provider  # type: ignore[attr-defined]
+    if model is not None:
+        exc.model = model  # type: ignore[attr-defined]
+    return exc
+
+
+class TestLiteLLMErrorExtraction:
+    """Verify that LiteLLM error attributes are included in the ValueError."""
+
+    def test_status_code_included(self) -> None:
+        mock_llm = MagicMock()
+        mock_llm.invoke.side_effect = _make_litellm_style_error(
+            message="Content filter triggered",
+            status_code=400,
+            llm_provider="azure",
+            model="gpt-4o",
+        )
+
+        with pytest.raises(ValueError, match="status_code=400"):
+            _summarize_image("data:image/png;base64,abc", mock_llm)
+
+    def test_llm_provider_included(self) -> None:
+        mock_llm = MagicMock()
+        mock_llm.invoke.side_effect = _make_litellm_style_error(
+            message="Bad request",
+            status_code=400,
+            llm_provider="azure",
+        )
+
+        with pytest.raises(ValueError, match="llm_provider=azure"):
+            _summarize_image("data:image/png;base64,abc", mock_llm)
+
+    def test_model_included(self) -> None:
+        mock_llm = MagicMock()
+        mock_llm.invoke.side_effect = _make_litellm_style_error(
+            message="Bad request",
+            model="gpt-4o",
+        )
+
+        with pytest.raises(ValueError, match="model=gpt-4o"):
+            _summarize_image("data:image/png;base64,abc", mock_llm)
+
+    def test_all_fields_in_single_message(self) -> None:
+        mock_llm = MagicMock()
+        mock_llm.invoke.side_effect = _make_litellm_style_error(
+            message="Rate limit exceeded",
+            status_code=429,
+            llm_provider="azure",
+            model="gpt-4o",
+        )
+
+        with pytest.raises(ValueError) as exc_info:
+            _summarize_image("data:image/png;base64,abc", mock_llm)
+
+        msg = str(exc_info.value)
+        assert "status_code=429" in msg
+        assert "llm_provider=azure" in msg
+        assert "model=gpt-4o" in msg
+        assert "Rate limit exceeded" in msg
+
+    def test_plain_exception_without_litellm_attrs(self) -> None:
+        """Non-LiteLLM exceptions should still produce a useful message."""
+        mock_llm = MagicMock()
+        mock_llm.invoke.side_effect = ConnectionError("Connection refused")
+
+        with pytest.raises(ValueError) as exc_info:
+            _summarize_image("data:image/png;base64,abc", mock_llm)
+
+        msg = str(exc_info.value)
+        assert "ConnectionError" in msg
+        assert "Connection refused" in msg
+        # Should not contain status_code/llm_provider/model
+        assert "status_code" not in msg
+        assert "llm_provider" not in msg
+
+    def test_no_base64_in_error(self) -> None:
+        """Error messages must not contain the full base64 image payload.
+
+        Some LiteLLM exceptions echo the request body (including base64 images)
+        in their message.  The truncation guard ensures the bulk of such a
+        payload is stripped from the re-raised ValueError.
+        """
+        mock_llm = MagicMock()
+        # Build a long base64-like payload that exceeds the 512-char truncation
+        fake_b64_payload = "iVBORw0KGgo" * 100  # ~1100 chars
+        fake_b64 = f"data:image/png;base64,{fake_b64_payload}"
+
+        mock_llm.invoke.side_effect = RuntimeError(
+            f"Request failed for payload: {fake_b64}"
+        )
+
+        with pytest.raises(ValueError) as exc_info:
+            _summarize_image(fake_b64, mock_llm)
+
+        msg = str(exc_info.value)
+        # The full payload must not appear (truncation should have kicked in)
+        assert fake_b64_payload not in msg
+        assert "truncated" in msg
+
+    def test_long_error_message_truncated(self) -> None:
+        """Exception messages longer than 512 chars are truncated."""
+        mock_llm = MagicMock()
+        long_msg = "x" * 1000
+        mock_llm.invoke.side_effect = RuntimeError(long_msg)
+
+        with pytest.raises(ValueError) as exc_info:
+            _summarize_image("data:image/png;base64,abc", mock_llm)
+
+        msg = str(exc_info.value)
+        assert "truncated" in msg
+        # The full 1000-char string should not appear
+        assert long_msg not in msg

backend/tests/unit/onyx/llm/test_multi_llm.py

@@ -28,6 +28,7 @@ from onyx.llm.utils import get_max_input_tokens
 VERTEX_OPUS_MODELS_REJECTING_OUTPUT_CONFIG = [
     "claude-opus-4-5@20251101",
     "claude-opus-4-6",
+    "claude-opus-4-7",
 ]
 
 

backend/tests/unit/onyx/server/manage/llm/test_fetch_models_api.py

@@ -503,6 +503,7 @@ class TestGetLMStudioAvailableModels:
 
         mock_session = MagicMock()
         mock_provider = MagicMock()
+        mock_provider.api_base = "http://localhost:1234"
         mock_provider.custom_config = {"LM_STUDIO_API_KEY": "stored-secret"}
 
         response = {

backend/tests/unit/onyx/server/manage/llm/test_llm_provider_utils.py

@@ -3,6 +3,7 @@
 from onyx.server.manage.llm.utils import generate_bedrock_display_name
 from onyx.server.manage.llm.utils import generate_ollama_display_name
 from onyx.server.manage.llm.utils import infer_vision_support
+from onyx.server.manage.llm.utils import is_embedding_model
 from onyx.server.manage.llm.utils import is_reasoning_model
 from onyx.server.manage.llm.utils import is_valid_bedrock_model
 from onyx.server.manage.llm.utils import strip_openrouter_vendor_prefix
@@ -209,3 +210,35 @@ class TestIsReasoningModel:
             is_reasoning_model("anthropic/claude-3-5-sonnet", "Claude 3.5 Sonnet")
             is False
         )
+
+
+class TestIsEmbeddingModel:
+    """Tests for embedding model detection."""
+
+    def test_openai_embedding_ada(self) -> None:
+        assert is_embedding_model("text-embedding-ada-002") is True
+
+    def test_openai_embedding_3_small(self) -> None:
+        assert is_embedding_model("text-embedding-3-small") is True
+
+    def test_openai_embedding_3_large(self) -> None:
+        assert is_embedding_model("text-embedding-3-large") is True
+
+    def test_cohere_embed_model(self) -> None:
+        assert is_embedding_model("embed-english-v3.0") is True
+
+    def test_bedrock_titan_embed(self) -> None:
+        assert is_embedding_model("amazon.titan-embed-text-v1") is True
+
+    def test_gpt4o_not_embedding(self) -> None:
+        assert is_embedding_model("gpt-4o") is False
+
+    def test_gpt4_not_embedding(self) -> None:
+        assert is_embedding_model("gpt-4") is False
+
+    def test_dall_e_not_embedding(self) -> None:
+        assert is_embedding_model("dall-e-3") is False
+
+    def test_unknown_custom_model_not_embedding(self) -> None:
+        """Custom/local models not in litellm's model DB should default to False."""
+        assert is_embedding_model("my-custom-local-model-v1") is False

backend/tests/unit/onyx/server/test_upload_files.py

@@ -0,0 +1,109 @@
+import io
+import zipfile
+from unittest.mock import MagicMock
+from unittest.mock import patch
+from zipfile import BadZipFile
+
+import pytest
+from fastapi import UploadFile
+from starlette.datastructures import Headers
+
+from onyx.configs.constants import FileOrigin
+from onyx.server.documents.connector import upload_files
+
+
+def _create_test_zip() -> bytes:
+    """Create a simple in-memory zip file containing two text files."""
+    buf = io.BytesIO()
+    with zipfile.ZipFile(buf, "w") as zf:
+        zf.writestr("file1.txt", "hello")
+        zf.writestr("file2.txt", "world")
+    return buf.getvalue()
+
+
+def _make_upload_file(content: bytes, filename: str, content_type: str) -> UploadFile:
+    return UploadFile(
+        file=io.BytesIO(content),
+        filename=filename,
+        headers=Headers({"content-type": content_type}),
+    )
+
+
+@patch("onyx.server.documents.connector.get_default_file_store")
+def test_upload_zip_with_unzip_true_extracts_files(
+    mock_get_store: MagicMock,
+) -> None:
+    """When unzip=True (default), a zip upload is extracted into individual files."""
+    mock_store = MagicMock()
+    mock_store.save_file.side_effect = lambda **kwargs: f"id-{kwargs['display_name']}"
+    mock_get_store.return_value = mock_store
+
+    zip_bytes = _create_test_zip()
+    upload = _make_upload_file(zip_bytes, "test.zip", "application/zip")
+
+    result = upload_files([upload], FileOrigin.CONNECTOR)
+
+    # Should have extracted the two individual files, not stored the zip itself
+    assert len(result.file_paths) == 2
+    assert "id-file1.txt" in result.file_paths
+    assert "id-file2.txt" in result.file_paths
+    assert "file1.txt" in result.file_names
+    assert "file2.txt" in result.file_names
+
+
+@patch("onyx.server.documents.connector.get_default_file_store")
+def test_upload_zip_with_unzip_false_stores_zip_as_is(
+    mock_get_store: MagicMock,
+) -> None:
+    """When unzip=False, the zip file is stored as-is without extraction."""
+    mock_store = MagicMock()
+    mock_store.save_file.return_value = "zip-file-id"
+    mock_get_store.return_value = mock_store
+
+    zip_bytes = _create_test_zip()
+    upload = _make_upload_file(zip_bytes, "site_export.zip", "application/zip")
+
+    result = upload_files([upload], FileOrigin.CONNECTOR, unzip=False)
+
+    # Should store exactly one file (the zip itself)
+    assert len(result.file_paths) == 1
+    assert result.file_paths[0] == "zip-file-id"
+    assert result.file_names == ["site_export.zip"]
+    # No zip metadata should be created
+    assert result.zip_metadata_file_id is None
+
+    # Verify the stored content is a valid zip
+    saved_content: io.BytesIO = mock_store.save_file.call_args[1]["content"]
+    saved_content.seek(0)
+    with zipfile.ZipFile(saved_content, "r") as zf:
+        assert set(zf.namelist()) == {"file1.txt", "file2.txt"}
+
+
+@patch("onyx.server.documents.connector.get_default_file_store")
+def test_upload_invalid_zip_with_unzip_false_raises(
+    mock_get_store: MagicMock,
+) -> None:
+    """An invalid zip is rejected even when unzip=False (validation still runs)."""
+    mock_get_store.return_value = MagicMock()
+
+    bad_zip = _make_upload_file(b"not a zip", "bad.zip", "application/zip")
+
+    with pytest.raises(BadZipFile):
+        upload_files([bad_zip], FileOrigin.CONNECTOR, unzip=False)
+
+
+@patch("onyx.server.documents.connector.get_default_file_store")
+def test_upload_multiple_zips_rejected_when_unzip_false(
+    mock_get_store: MagicMock,
+) -> None:
+    """The seen_zip guard rejects a second zip even when unzip=False."""
+    mock_store = MagicMock()
+    mock_store.save_file.return_value = "zip-id"
+    mock_get_store.return_value = mock_store
+
+    zip_bytes = _create_test_zip()
+    zip1 = _make_upload_file(zip_bytes, "a.zip", "application/zip")
+    zip2 = _make_upload_file(zip_bytes, "b.zip", "application/zip")
+
+    with pytest.raises(Exception, match="Only one zip file"):
+        upload_files([zip1, zip2], FileOrigin.CONNECTOR, unzip=False)

backend/tests/unit/onyx/tools/tool_implementations/python/test_python_tool_upload_cache.py

@@ -0,0 +1,208 @@
+"""Unit tests for PythonTool file-upload caching.
+
+Verifies that PythonTool reuses code-interpreter file IDs across multiple
+run() calls within the same session instead of re-uploading identical content
+on every agent loop iteration.
+"""
+
+from unittest.mock import MagicMock
+from unittest.mock import patch
+
+from onyx.tools.models import ChatFile
+from onyx.tools.models import PythonToolOverrideKwargs
+from onyx.tools.tool_implementations.python.code_interpreter_client import (
+    StreamResultEvent,
+)
+from onyx.tools.tool_implementations.python.python_tool import PythonTool
+
+TOOL_MODULE = "onyx.tools.tool_implementations.python.python_tool"
+
+
+def _make_stream_result() -> StreamResultEvent:
+    return StreamResultEvent(
+        exit_code=0,
+        timed_out=False,
+        duration_ms=10,
+        files=[],
+    )
+
+
+def _make_tool() -> PythonTool:
+    emitter = MagicMock()
+    return PythonTool(tool_id=1, emitter=emitter)
+
+
+def _make_override(files: list[ChatFile]) -> PythonToolOverrideKwargs:
+    return PythonToolOverrideKwargs(chat_files=files)
+
+
+def _run_tool(tool: PythonTool, mock_client: MagicMock, files: list[ChatFile]) -> None:
+    """Call tool.run() with a mocked CodeInterpreterClient context manager."""
+    from onyx.server.query_and_chat.placement import Placement
+
+    mock_client.execute_streaming.return_value = iter([_make_stream_result()])
+
+    ctx = MagicMock()
+    ctx.__enter__ = MagicMock(return_value=mock_client)
+    ctx.__exit__ = MagicMock(return_value=False)
+
+    placement = Placement(turn_index=0, tab_index=0)
+    override = _make_override(files)
+
+    with patch(f"{TOOL_MODULE}.CodeInterpreterClient", return_value=ctx):
+        tool.run(placement=placement, override_kwargs=override, code="print('hi')")
+
+
+# ---------------------------------------------------------------------------
+# Cache hit: same content uploaded in a second call reuses the file_id
+# ---------------------------------------------------------------------------
+
+
+@patch(f"{TOOL_MODULE}.CODE_INTERPRETER_BASE_URL", "http://fake:8000")
+def test_same_file_uploaded_only_once_across_two_runs() -> None:
+    tool = _make_tool()
+    client = MagicMock()
+    client.upload_file.return_value = "file-id-abc"
+
+    pptx_content = b"fake pptx bytes"
+    files = [ChatFile(filename="report.pptx", content=pptx_content)]
+
+    _run_tool(tool, client, files)
+    _run_tool(tool, client, files)
+
+    # upload_file should only have been called once across both runs
+    client.upload_file.assert_called_once_with(pptx_content, "report.pptx")
+
+
+@patch(f"{TOOL_MODULE}.CODE_INTERPRETER_BASE_URL", "http://fake:8000")
+def test_cached_file_id_is_staged_on_second_run() -> None:
+    tool = _make_tool()
+    client = MagicMock()
+    client.upload_file.return_value = "file-id-abc"
+
+    files = [ChatFile(filename="data.pptx", content=b"content")]
+
+    _run_tool(tool, client, files)
+
+    # On the second run, execute_streaming should still receive the file
+    client.execute_streaming.return_value = iter([_make_stream_result()])
+    ctx = MagicMock()
+    ctx.__enter__ = MagicMock(return_value=client)
+    ctx.__exit__ = MagicMock(return_value=False)
+
+    from onyx.server.query_and_chat.placement import Placement
+
+    placement = Placement(turn_index=1, tab_index=0)
+    with patch(f"{TOOL_MODULE}.CodeInterpreterClient", return_value=ctx):
+        tool.run(
+            placement=placement,
+            override_kwargs=_make_override(files),
+            code="print('hi')",
+        )
+
+    # The second execute_streaming call should include the file
+    _, kwargs = client.execute_streaming.call_args
+    staged_files = kwargs.get("files") or []
+    assert any(f["file_id"] == "file-id-abc" for f in staged_files)
+
+
+# ---------------------------------------------------------------------------
+# Cache miss: different content triggers a new upload
+# ---------------------------------------------------------------------------
+
+
+@patch(f"{TOOL_MODULE}.CODE_INTERPRETER_BASE_URL", "http://fake:8000")
+def test_different_file_content_uploaded_separately() -> None:
+    tool = _make_tool()
+    client = MagicMock()
+    client.upload_file.side_effect = ["file-id-v1", "file-id-v2"]
+
+    file_v1 = ChatFile(filename="report.pptx", content=b"version 1")
+    file_v2 = ChatFile(filename="report.pptx", content=b"version 2")
+
+    _run_tool(tool, client, [file_v1])
+    _run_tool(tool, client, [file_v2])
+
+    assert client.upload_file.call_count == 2
+
+
+@patch(f"{TOOL_MODULE}.CODE_INTERPRETER_BASE_URL", "http://fake:8000")
+def test_multiple_distinct_files_each_uploaded_once() -> None:
+    tool = _make_tool()
+    client = MagicMock()
+    client.upload_file.side_effect = ["id-a", "id-b"]
+
+    files = [
+        ChatFile(filename="a.pptx", content=b"aaa"),
+        ChatFile(filename="b.xlsx", content=b"bbb"),
+    ]
+
+    _run_tool(tool, client, files)
+    _run_tool(tool, client, files)
+
+    # Two distinct files — each uploaded exactly once
+    assert client.upload_file.call_count == 2
+
+
+@patch(f"{TOOL_MODULE}.CODE_INTERPRETER_BASE_URL", "http://fake:8000")
+def test_same_content_different_filename_uploaded_separately() -> None:
+    # Identical bytes but different names must each get their own upload slot
+    # so both files appear under their respective paths in the workspace.
+    tool = _make_tool()
+    client = MagicMock()
+    client.upload_file.side_effect = ["id-v1", "id-v2"]
+
+    same_bytes = b"shared content"
+    files = [
+        ChatFile(filename="report_v1.csv", content=same_bytes),
+        ChatFile(filename="report_v2.csv", content=same_bytes),
+    ]
+
+    _run_tool(tool, client, files)
+
+    assert client.upload_file.call_count == 2
+
+
+# ---------------------------------------------------------------------------
+# No cross-instance sharing: a fresh PythonTool re-uploads everything
+# ---------------------------------------------------------------------------
+
+
+@patch(f"{TOOL_MODULE}.CODE_INTERPRETER_BASE_URL", "http://fake:8000")
+def test_new_tool_instance_re_uploads_file() -> None:
+    client = MagicMock()
+    client.upload_file.side_effect = ["id-session-1", "id-session-2"]
+
+    files = [ChatFile(filename="deck.pptx", content=b"slide data")]
+
+    tool_session_1 = _make_tool()
+    _run_tool(tool_session_1, client, files)
+
+    tool_session_2 = _make_tool()
+    _run_tool(tool_session_2, client, files)
+
+    # Different instances — each uploads independently
+    assert client.upload_file.call_count == 2
+
+
+# ---------------------------------------------------------------------------
+# Upload failure: failed upload is not cached, retried next run
+# ---------------------------------------------------------------------------
+
+
+@patch(f"{TOOL_MODULE}.CODE_INTERPRETER_BASE_URL", "http://fake:8000")
+def test_upload_failure_not_cached() -> None:
+    tool = _make_tool()
+    client = MagicMock()
+    # First call raises, second succeeds
+    client.upload_file.side_effect = [Exception("network error"), "file-id-ok"]
+
+    files = [ChatFile(filename="slides.pptx", content=b"data")]
+
+    # First run — upload fails, file is skipped but not cached
+    _run_tool(tool, client, files)
+
+    # Second run — should attempt upload again
+    _run_tool(tool, client, files)
+
+    assert client.upload_file.call_count == 2

backend/tests/unit/onyx/utils/test_threadpool_concurrency.py

@@ -32,15 +32,17 @@ def test_run_with_timeout_raises_on_timeout(slow: float, timeout: float) -> None
     """Test that a function that exceeds timeout raises TimeoutError"""
 
     def slow_function() -> None:
-        time.sleep(slow)  # Sleep for 2 seconds
+        time.sleep(slow)
 
+    start = time.monotonic()
     with pytest.raises(TimeoutError) as exc_info:
-        start = time.time()
-        run_with_timeout(timeout, slow_function)  # Set timeout to 0.1 seconds
-        end = time.time()
-        assert end - start >= timeout
-        assert end - start < (slow + timeout) / 2
+        run_with_timeout(timeout, slow_function)
+    elapsed = time.monotonic() - start
+
     assert f"timed out after {timeout} seconds" in str(exc_info.value)
+    assert elapsed >= timeout
+    # Should return around the timeout duration, not the full sleep duration
+    assert elapsed == pytest.approx(timeout, abs=0.8)
 
 
 @pytest.mark.filterwarnings("ignore::pytest.PytestUnhandledThreadExceptionWarning")

backend/tests/unit/onyx/utils/test_vespa_query.py

@@ -44,13 +44,13 @@ class TestBuildVespaFilters:
         assert result == f'({SOURCE_TYPE} contains "web") and '
 
     def test_acl(self) -> None:
-        """Test with acls."""
+        """Test with acls — uses weightedSet operator for efficient matching."""
         # Single ACL
         filters = IndexFilters(access_control_list=["user1"])
         result = build_vespa_filters(filters)
         assert (
             result
-            == f'!({HIDDEN}=true) and (access_control_list contains "user1") and '
+            == f'!({HIDDEN}=true) and weightedSet(access_control_list, {{"user1":1}}) and '
         )
 
         # Multiple ACL's
@@ -58,7 +58,7 @@ class TestBuildVespaFilters:
         result = build_vespa_filters(filters)
         assert (
             result
-            == f'!({HIDDEN}=true) and (access_control_list contains "user2" or access_control_list contains "group2") and '
+            == f'!({HIDDEN}=true) and weightedSet(access_control_list, {{"user2":1, "group2":1}}) and '
         )
 
     def test_tenant_filter(self) -> None:
@@ -250,10 +250,7 @@ class TestBuildVespaFilters:
         result = build_vespa_filters(filters)
 
         expected = f"!({HIDDEN}=true) and "
-        expected += (
-            '(access_control_list contains "user1" or '
-            'access_control_list contains "group1") and '
-        )
+        expected += 'weightedSet(access_control_list, {"user1":1, "group1":1}) and '
         expected += f'({SOURCE_TYPE} contains "web") and '
         expected += f'({METADATA_LIST} contains "color{INDEX_SEPARATOR}red") and '
         # Knowledge scope filters are OR'd together
@@ -298,6 +295,38 @@ class TestBuildVespaFilters:
         )
         assert expected == result
 
+    def test_acl_large_list_uses_weighted_set(self) -> None:
+        """Verify that large ACL lists produce a weightedSet clause
+        instead of OR-chained contains — this is what prevents Vespa
+        HTTP 400 errors for users with thousands of permission groups."""
+        acl = [f"external_group:google_drive_{i}" for i in range(10_000)]
+        acl += ["user_email:user@example.com", "__PUBLIC__"]
+        filters = IndexFilters(access_control_list=acl)
+        result = build_vespa_filters(filters)
+
+        assert "weightedSet(access_control_list, {" in result
+        # Must NOT contain OR-chained contains clauses
+        assert "access_control_list contains" not in result
+        # All entries should be present
+        assert '"external_group:google_drive_0":1' in result
+        assert '"external_group:google_drive_9999":1' in result
+        assert '"user_email:user@example.com":1' in result
+        assert '"__PUBLIC__":1' in result
+
+    def test_acl_empty_strings_filtered(self) -> None:
+        """Empty strings in the ACL list should be filtered out."""
+        filters = IndexFilters(access_control_list=["user1", "", "group1"])
+        result = build_vespa_filters(filters)
+        assert (
+            result
+            == f'!({HIDDEN}=true) and weightedSet(access_control_list, {{"user1":1, "group1":1}}) and '
+        )
+
+        # All empty
+        filters = IndexFilters(access_control_list=["", ""])
+        result = build_vespa_filters(filters)
+        assert result == f"!({HIDDEN}=true) and "
+
     def test_empty_or_none_values(self) -> None:
         """Test with empty or None values in filter lists."""
         # Empty strings in document set

pyproject.toml

@@ -12,7 +12,7 @@ dependencies = [
     "cohere==5.6.1",
     "fastapi==0.133.1",
     "google-genai==1.52.0",
-    "litellm==1.81.6",
+    "litellm[google]==1.81.6",
     "openai==2.14.0",
     "pydantic==2.11.7",
     "prometheus_client>=0.21.1",
@@ -237,7 +237,7 @@ line-length = 130
 target-version = "py311"
 
 [tool.ruff.lint]
-ignore = []
+ignore = ["E501"]
 select = [
   "ARG",
   "E",

tools/ods/cmd/print_latest.go

@@ -0,0 +1,35 @@
+package cmd
+
+import (
+	"fmt"
+	"github.com/jmelahman/tag/git"
+	"github.com/spf13/cobra"
+)
+
+// NewLatestStableTagCommand creates the latest-stable-tag command.
+func NewLatestStableTagCommand() *cobra.Command {
+	cmd := &cobra.Command{
+		Use:   "latest-stable-tag",
+		Short: "Print the git tag that should receive the 'latest' Docker tag",
+		Long: `Print the highest stable (non-pre-release) semver tag in the repository.
+
+This is used during deployment to decide whether a given tag should
+receive the "latest" tag on Docker Hub. Only the highest vX.Y.Z tag
+qualifies. Tags with pre-release suffixes (e.g. v1.2.3-beta,
+v1.2.3-cloud.1) are excluded.`,
+		Args: cobra.NoArgs,
+		RunE: func(c *cobra.Command, _ []string) error {
+			tag, err := git.GetLatestStableSemverTag("")
+			if err != nil {
+				return fmt.Errorf("get latest stable semver tag: %w", err)
+			}
+			if tag == "" {
+				return fmt.Errorf("no stable semver tag found in repository")
+			}
+			fmt.Println(tag)
+			return nil
+		},
+	}
+
+	return cmd
+}

tools/ods/cmd/root.go

@@ -52,6 +52,7 @@ func NewRootCommand() *cobra.Command {
 	cmd.AddCommand(NewScreenshotDiffCommand())
 	cmd.AddCommand(NewDesktopCommand())
 	cmd.AddCommand(NewWebCommand())
+	cmd.AddCommand(NewLatestStableTagCommand())
 	cmd.AddCommand(NewWhoisCommand())
 
 	return cmd

tools/ods/go.mod

@@ -3,12 +3,13 @@ module github.com/onyx-dot-app/onyx/tools/ods
 go 1.26.0
 
 require (
+	github.com/jmelahman/tag v0.5.2
 	github.com/sirupsen/logrus v1.9.3
-	github.com/spf13/cobra v1.10.1
-	github.com/spf13/pflag v1.0.9
+	github.com/spf13/cobra v1.10.2
+	github.com/spf13/pflag v1.0.10
 )
 
 require (
 	github.com/inconshreveable/mousetrap v1.1.0 // indirect
-	golang.org/x/sys v0.0.0-20220715151400-c0bba94af5f8 // indirect
+	golang.org/x/sys v0.39.0 // indirect
 )

tools/ods/go.sum

@@ -4,20 +4,26 @@ github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/inconshreveable/mousetrap v1.1.0 h1:wN+x4NVGpMsO7ErUn/mUI3vEoE6Jt13X2s0bqwp9tc8=
 github.com/inconshreveable/mousetrap v1.1.0/go.mod h1:vpF70FUmC8bwa3OWnCshd2FqLfsEA9PFc4w1p2J65bw=
+github.com/jmelahman/tag v0.5.2 h1:g6A/aHehu5tkA31mPoDsXBNr1FigZ9A82Y8WVgb/WsM=
+github.com/jmelahman/tag v0.5.2/go.mod h1:qmuqk19B1BKkpcg3kn7l/Eey+UqucLxgOWkteUGiG4Q=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/russross/blackfriday/v2 v2.1.0/go.mod h1:+Rmxgy9KzJVeS9/2gXHxylqXiyQDYRxCVz55jmeOWTM=
 github.com/sirupsen/logrus v1.9.3 h1:dueUQJ1C2q9oE3F7wvmSGAaVtTmUizReu6fjN8uqzbQ=
 github.com/sirupsen/logrus v1.9.3/go.mod h1:naHLuLoDiP4jHNo9R0sCBMtWGeIprob74mVsIT4qYEQ=
-github.com/spf13/cobra v1.10.1 h1:lJeBwCfmrnXthfAupyUTzJ/J4Nc1RsHC/mSRU2dll/s=
-github.com/spf13/cobra v1.10.1/go.mod h1:7SmJGaTHFVBY0jW4NXGluQoLvhqFQM+6XSKD+P4XaB0=
-github.com/spf13/pflag v1.0.9 h1:9exaQaMOCwffKiiiYk6/BndUBv+iRViNW+4lEMi0PvY=
+github.com/spf13/cobra v1.10.2 h1:DMTTonx5m65Ic0GOoRY2c16WCbHxOOw6xxezuLaBpcU=
+github.com/spf13/cobra v1.10.2/go.mod h1:7C1pvHqHw5A4vrJfjNwvOdzYu0Gml16OCs2GRiTUUS4=
 github.com/spf13/pflag v1.0.9/go.mod h1:McXfInJRrz4CZXVZOBLb0bTZqETkiAhM9Iw0y3An2Bg=
+github.com/spf13/pflag v1.0.10 h1:4EBh2KAYBwaONj6b2Ye1GiHfwjqyROoF4RwYO+vPwFk=
+github.com/spf13/pflag v1.0.10/go.mod h1:McXfInJRrz4CZXVZOBLb0bTZqETkiAhM9Iw0y3An2Bg=
 github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
-github.com/stretchr/testify v1.7.0 h1:nwc3DEeHmmLAfoZucVR881uASk0Mfjw8xYJ99tb5CcY=
 github.com/stretchr/testify v1.7.0/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
-golang.org/x/sys v0.0.0-20220715151400-c0bba94af5f8 h1:0A+M6Uqn+Eje4kHMK80dtF3JCXC4ykBgQG4Fe06QRhQ=
+github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
+github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
+go.yaml.in/yaml/v3 v3.0.4/go.mod h1:DhzuOOF2ATzADvBadXxruRBLzYTpT36CKvDb3+aBEFg=
 golang.org/x/sys v0.0.0-20220715151400-c0bba94af5f8/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
+golang.org/x/sys v0.39.0 h1:CvCKL8MeisomCi6qNZ+wbb0DN9E5AATixKsvNtMoMFk=
+golang.org/x/sys v0.39.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
 gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=

uv.lock

@@ -2107,6 +2107,12 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/ed/d4/90197b416cb61cefd316964fd9e7bd8324bcbafabf40eef14a9f20b81974/google_api_core-2.28.1-py3-none-any.whl", hash = "sha256:4021b0f8ceb77a6fb4de6fde4502cecab45062e66ff4f2895169e0b35bc9466c", size = 173706, upload-time = "2025-10-28T21:34:50.151Z" },
 ]
 
+[package.optional-dependencies]
+grpc = [
+    { name = "grpcio" },
+    { name = "grpcio-status" },
+]
+
 [[package]]
 name = "google-api-python-client"
 version = "2.86.0"
@@ -2164,6 +2170,124 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/4a/07/8d9a8186e6768b55dfffeb57c719bc03770cf8a970a074616ae6f9e26a57/google_auth_oauthlib-1.0.0-py2.py3-none-any.whl", hash = "sha256:95880ca704928c300f48194d1770cf5b1462835b6e49db61445a520f793fd5fb", size = 18926, upload-time = "2023-02-07T20:53:18.837Z" },
 ]
 
+[[package]]
+name = "google-cloud-aiplatform"
+version = "1.133.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "docstring-parser" },
+    { name = "google-api-core", extra = ["grpc"] },
+    { name = "google-auth" },
+    { name = "google-cloud-bigquery" },
+    { name = "google-cloud-resource-manager" },
+    { name = "google-cloud-storage" },
+    { name = "google-genai" },
+    { name = "packaging" },
+    { name = "proto-plus" },
+    { name = "protobuf" },
+    { name = "pydantic" },
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/d4/be/31ce7fd658ddebafbe5583977ddee536b2bacc491ad10b5a067388aec66f/google_cloud_aiplatform-1.133.0.tar.gz", hash = "sha256:3a6540711956dd178daaab3c2c05db476e46d94ac25912b8cf4f59b00b058ae0", size = 9921309, upload-time = "2026-01-08T22:11:25.079Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/01/5b/ef74ff65aebb74eaba51078e33ddd897247ba0d1197fd5a7953126205519/google_cloud_aiplatform-1.133.0-py2.py3-none-any.whl", hash = "sha256:dfc81228e987ca10d1c32c7204e2131b3c8d6b7c8e0b4e23bf7c56816bc4c566", size = 8184595, upload-time = "2026-01-08T22:11:22.067Z" },
+]
+
+[[package]]
+name = "google-cloud-bigquery"
+version = "3.41.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "google-api-core", extra = ["grpc"] },
+    { name = "google-auth" },
+    { name = "google-cloud-core" },
+    { name = "google-resumable-media" },
+    { name = "packaging" },
+    { name = "python-dateutil" },
+    { name = "requests" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/ce/13/6515c7aab55a4a0cf708ffd309fb9af5bab54c13e32dc22c5acd6497193c/google_cloud_bigquery-3.41.0.tar.gz", hash = "sha256:2217e488b47ed576360c9b2cc07d59d883a54b83167c0ef37f915c26b01a06fe", size = 513434, upload-time = "2026-03-30T22:50:55.347Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/40/33/1d3902efadef9194566d499d61507e1f038454e0b55499d2d7f8ab2a4fee/google_cloud_bigquery-3.41.0-py3-none-any.whl", hash = "sha256:2a5b5a737b401cbd824a6e5eac7554100b878668d908e6548836b5d8aaa4dcaa", size = 262343, upload-time = "2026-03-30T22:48:45.444Z" },
+]
+
+[[package]]
+name = "google-cloud-core"
+version = "2.5.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "google-api-core" },
+    { name = "google-auth" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/dc/24/6ca08b0a03c7b0c620427503ab00353a4ae806b848b93bcea18b6b76fde6/google_cloud_core-2.5.1.tar.gz", hash = "sha256:3dc94bdec9d05a31d9f355045ed0f369fbc0d8c665076c734f065d729800f811", size = 36078, upload-time = "2026-03-30T22:50:08.057Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/73/d9/5bb050cb32826466aa9b25f79e2ca2879fe66cb76782d4ed798dd7506151/google_cloud_core-2.5.1-py3-none-any.whl", hash = "sha256:ea62cdf502c20e3e14be8a32c05ed02113d7bef454e40ff3fab6fe1ec9f1f4e7", size = 29452, upload-time = "2026-03-30T22:48:31.567Z" },
+]
+
+[[package]]
+name = "google-cloud-resource-manager"
+version = "1.17.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "google-api-core", extra = ["grpc"] },
+    { name = "google-auth" },
+    { name = "grpc-google-iam-v1" },
+    { name = "grpcio" },
+    { name = "proto-plus" },
+    { name = "protobuf" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b2/1a/13060cabf553d52d151d2afc26b39561e82853380d499dd525a0d422d9f0/google_cloud_resource_manager-1.17.0.tar.gz", hash = "sha256:0f486b62e2c58ff992a3a50fa0f4a96eef7750aa6c971bb373398ccb91828660", size = 464971, upload-time = "2026-03-26T22:17:29.204Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/3d/f7/661d7a9023e877a226b5683429c3662f75a29ef45cb1464cf39adb689218/google_cloud_resource_manager-1.17.0-py3-none-any.whl", hash = "sha256:e479baf4b014a57f298e01b8279e3290b032e3476d69c8e5e1427af8f82739a5", size = 404403, upload-time = "2026-03-26T22:15:26.57Z" },
+]
+
+[[package]]
+name = "google-cloud-storage"
+version = "3.10.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "google-api-core" },
+    { name = "google-auth" },
+    { name = "google-cloud-core" },
+    { name = "google-crc32c" },
+    { name = "google-resumable-media" },
+    { name = "requests" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/4c/47/205eb8e9a1739b5345843e5a425775cbdc472cc38e7eda082ba5b8d02450/google_cloud_storage-3.10.1.tar.gz", hash = "sha256:97db9aa4460727982040edd2bd13ff3d5e2260b5331ad22895802da1fc2a5286", size = 17309950, upload-time = "2026-03-23T09:35:23.409Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/ad/ff/ca9ab2417fa913d75aae38bf40bf856bb2749a604b2e0f701b37cfcd23cc/google_cloud_storage-3.10.1-py3-none-any.whl", hash = "sha256:a72f656759b7b99bda700f901adcb3425a828d4a29f911bc26b3ea79c5b1217f", size = 324453, upload-time = "2026-03-23T09:35:21.368Z" },
+]
+
+[[package]]
+name = "google-crc32c"
+version = "1.8.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/03/41/4b9c02f99e4c5fb477122cd5437403b552873f014616ac1d19ac8221a58d/google_crc32c-1.8.0.tar.gz", hash = "sha256:a428e25fb7691024de47fecfbff7ff957214da51eddded0da0ae0e0f03a2cf79", size = 14192, upload-time = "2025-12-16T00:35:25.142Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/5d/ef/21ccfaab3d5078d41efe8612e0ed0bfc9ce22475de074162a91a25f7980d/google_crc32c-1.8.0-cp311-cp311-macosx_12_0_arm64.whl", hash = "sha256:014a7e68d623e9a4222d663931febc3033c5c7c9730785727de2a81f87d5bab8", size = 31298, upload-time = "2025-12-16T00:20:32.241Z" },
+    { url = "https://files.pythonhosted.org/packages/c5/b8/f8413d3f4b676136e965e764ceedec904fe38ae8de0cdc52a12d8eb1096e/google_crc32c-1.8.0-cp311-cp311-macosx_12_0_x86_64.whl", hash = "sha256:86cfc00fe45a0ac7359e5214a1704e51a99e757d0272554874f419f79838c5f7", size = 30872, upload-time = "2025-12-16T00:33:58.785Z" },
+    { url = "https://files.pythonhosted.org/packages/f6/fd/33aa4ec62b290477181c55bb1c9302c9698c58c0ce9a6ab4874abc8b0d60/google_crc32c-1.8.0-cp311-cp311-manylinux1_x86_64.manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_5_x86_64.whl", hash = "sha256:19b40d637a54cb71e0829179f6cb41835f0fbd9e8eb60552152a8b52c36cbe15", size = 33243, upload-time = "2025-12-16T00:40:21.46Z" },
+    { url = "https://files.pythonhosted.org/packages/71/03/4820b3bd99c9653d1a5210cb32f9ba4da9681619b4d35b6a052432df4773/google_crc32c-1.8.0-cp311-cp311-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:17446feb05abddc187e5441a45971b8394ea4c1b6efd88ab0af393fd9e0a156a", size = 33608, upload-time = "2025-12-16T00:40:22.204Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/43/acf61476a11437bf9733fb2f70599b1ced11ec7ed9ea760fdd9a77d0c619/google_crc32c-1.8.0-cp311-cp311-win_amd64.whl", hash = "sha256:71734788a88f551fbd6a97be9668a0020698e07b2bf5b3aa26a36c10cdfb27b2", size = 34439, upload-time = "2025-12-16T00:35:20.458Z" },
+    { url = "https://files.pythonhosted.org/packages/e9/5f/7307325b1198b59324c0fa9807cafb551afb65e831699f2ce211ad5c8240/google_crc32c-1.8.0-cp312-cp312-macosx_12_0_arm64.whl", hash = "sha256:4b8286b659c1335172e39563ab0a768b8015e88e08329fa5321f774275fc3113", size = 31300, upload-time = "2025-12-16T00:21:56.723Z" },
+    { url = "https://files.pythonhosted.org/packages/21/8e/58c0d5d86e2220e6a37befe7e6a94dd2f6006044b1a33edf1ff6d9f7e319/google_crc32c-1.8.0-cp312-cp312-macosx_12_0_x86_64.whl", hash = "sha256:2a3dc3318507de089c5384cc74d54318401410f82aa65b2d9cdde9d297aca7cb", size = 30867, upload-time = "2025-12-16T00:38:31.302Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/a9/a780cc66f86335a6019f557a8aaca8fbb970728f0efd2430d15ff1beae0e/google_crc32c-1.8.0-cp312-cp312-manylinux1_x86_64.manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_5_x86_64.whl", hash = "sha256:14f87e04d613dfa218d6135e81b78272c3b904e2a7053b841481b38a7d901411", size = 33364, upload-time = "2025-12-16T00:40:22.96Z" },
+    { url = "https://files.pythonhosted.org/packages/21/3f/3457ea803db0198c9aaca2dd373750972ce28a26f00544b6b85088811939/google_crc32c-1.8.0-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:cb5c869c2923d56cb0c8e6bcdd73c009c36ae39b652dbe46a05eb4ef0ad01454", size = 33740, upload-time = "2025-12-16T00:40:23.96Z" },
+    { url = "https://files.pythonhosted.org/packages/df/c0/87c2073e0c72515bb8733d4eef7b21548e8d189f094b5dad20b0ecaf64f6/google_crc32c-1.8.0-cp312-cp312-win_amd64.whl", hash = "sha256:3cc0c8912038065eafa603b238abf252e204accab2a704c63b9e14837a854962", size = 34437, upload-time = "2025-12-16T00:35:21.395Z" },
+    { url = "https://files.pythonhosted.org/packages/d1/db/000f15b41724589b0e7bc24bc7a8967898d8d3bc8caf64c513d91ef1f6c0/google_crc32c-1.8.0-cp313-cp313-macosx_12_0_arm64.whl", hash = "sha256:3ebb04528e83b2634857f43f9bb8ef5b2bbe7f10f140daeb01b58f972d04736b", size = 31297, upload-time = "2025-12-16T00:23:20.709Z" },
+    { url = "https://files.pythonhosted.org/packages/d7/0d/8ebed0c39c53a7e838e2a486da8abb0e52de135f1b376ae2f0b160eb4c1a/google_crc32c-1.8.0-cp313-cp313-macosx_12_0_x86_64.whl", hash = "sha256:450dc98429d3e33ed2926fc99ee81001928d63460f8538f21a5d6060912a8e27", size = 30867, upload-time = "2025-12-16T00:43:14.628Z" },
+    { url = "https://files.pythonhosted.org/packages/ce/42/b468aec74a0354b34c8cbf748db20d6e350a68a2b0912e128cabee49806c/google_crc32c-1.8.0-cp313-cp313-manylinux1_x86_64.manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_5_x86_64.whl", hash = "sha256:3b9776774b24ba76831609ffbabce8cdf6fa2bd5e9df37b594221c7e333a81fa", size = 33344, upload-time = "2025-12-16T00:40:24.742Z" },
+    { url = "https://files.pythonhosted.org/packages/1c/e8/b33784d6fc77fb5062a8a7854e43e1e618b87d5ddf610a88025e4de6226e/google_crc32c-1.8.0-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:89c17d53d75562edfff86679244830599ee0a48efc216200691de8b02ab6b2b8", size = 33694, upload-time = "2025-12-16T00:40:25.505Z" },
+    { url = "https://files.pythonhosted.org/packages/92/b1/d3cbd4d988afb3d8e4db94ca953df429ed6db7282ed0e700d25e6c7bfc8d/google_crc32c-1.8.0-cp313-cp313-win_amd64.whl", hash = "sha256:57a50a9035b75643996fbf224d6661e386c7162d1dfdab9bc4ca790947d1007f", size = 34435, upload-time = "2025-12-16T00:35:22.107Z" },
+    { url = "https://files.pythonhosted.org/packages/21/88/8ecf3c2b864a490b9e7010c84fd203ec8cf3b280651106a3a74dd1b0ca72/google_crc32c-1.8.0-cp314-cp314-macosx_12_0_arm64.whl", hash = "sha256:e6584b12cb06796d285d09e33f63309a09368b9d806a551d8036a4207ea43697", size = 31301, upload-time = "2025-12-16T00:24:48.527Z" },
+    { url = "https://files.pythonhosted.org/packages/36/c6/f7ff6c11f5ca215d9f43d3629163727a272eabc356e5c9b2853df2bfe965/google_crc32c-1.8.0-cp314-cp314-macosx_12_0_x86_64.whl", hash = "sha256:f4b51844ef67d6cf2e9425983274da75f18b1597bb2c998e1c0a0e8d46f8f651", size = 30868, upload-time = "2025-12-16T00:48:12.163Z" },
+    { url = "https://files.pythonhosted.org/packages/56/15/c25671c7aad70f8179d858c55a6ae8404902abe0cdcf32a29d581792b491/google_crc32c-1.8.0-cp314-cp314-manylinux1_x86_64.manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_5_x86_64.whl", hash = "sha256:b0d1a7afc6e8e4635564ba8aa5c0548e3173e41b6384d7711a9123165f582de2", size = 33381, upload-time = "2025-12-16T00:40:26.268Z" },
+    { url = "https://files.pythonhosted.org/packages/42/fa/f50f51260d7b0ef5d4898af122d8a7ec5a84e2984f676f746445f783705f/google_crc32c-1.8.0-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:8b3f68782f3cbd1bce027e48768293072813469af6a61a86f6bb4977a4380f21", size = 33734, upload-time = "2025-12-16T00:40:27.028Z" },
+    { url = "https://files.pythonhosted.org/packages/08/a5/7b059810934a09fb3ccb657e0843813c1fee1183d3bc2c8041800374aa2c/google_crc32c-1.8.0-cp314-cp314-win_amd64.whl", hash = "sha256:d511b3153e7011a27ab6ee6bb3a5404a55b994dc1a7322c0b87b29606d9790e2", size = 34878, upload-time = "2025-12-16T00:35:23.142Z" },
+    { url = "https://files.pythonhosted.org/packages/52/c5/c171e4d8c44fec1422d801a6d2e5d7ddabd733eeda505c79730ee9607f07/google_crc32c-1.8.0-pp311-pypy311_pp73-manylinux1_x86_64.manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_5_x86_64.whl", hash = "sha256:87fa445064e7db928226b2e6f0d5304ab4cd0339e664a4e9a25029f384d9bb93", size = 28615, upload-time = "2025-12-16T00:40:29.298Z" },
+    { url = "https://files.pythonhosted.org/packages/9c/97/7d75fe37a7a6ed171a2cf17117177e7aab7e6e0d115858741b41e9dd4254/google_crc32c-1.8.0-pp311-pypy311_pp73-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:f639065ea2042d5c034bf258a9f085eaa7af0cd250667c0635a3118e8f92c69c", size = 28800, upload-time = "2025-12-16T00:40:30.322Z" },
+]
+
 [[package]]
 name = "google-genai"
 version = "1.52.0"
@@ -2183,6 +2307,18 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/ec/66/03f663e7bca7abe9ccfebe6cb3fe7da9a118fd723a5abb278d6117e7990e/google_genai-1.52.0-py3-none-any.whl", hash = "sha256:c8352b9f065ae14b9322b949c7debab8562982f03bf71d44130cd2b798c20743", size = 261219, upload-time = "2025-11-21T02:18:54.515Z" },
 ]
 
+[[package]]
+name = "google-resumable-media"
+version = "2.8.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "google-crc32c" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/3f/d1/b1ea14b93b6b78f57fc580125de44e9f593ab88dd2460f1a8a8d18f74754/google_resumable_media-2.8.2.tar.gz", hash = "sha256:f3354a182ebd193ae3f42e3ef95e6c9b10f128320de23ac7637236713b1acd70", size = 2164510, upload-time = "2026-03-30T23:34:25.369Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/5e/f8/50bfaf4658431ff9de45c5c3935af7ab01157a4903c603cd0eee6e78e087/google_resumable_media-2.8.2-py3-none-any.whl", hash = "sha256:82b6d8ccd11765268cdd2a2123f417ec806b8eef3000a9a38dfe3033da5fb220", size = 81511, upload-time = "2026-03-30T23:34:09.671Z" },
+]
+
 [[package]]
 name = "googleapis-common-protos"
 version = "1.72.0"
@@ -2195,6 +2331,11 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/c4/ab/09169d5a4612a5f92490806649ac8d41e3ec9129c636754575b3553f4ea4/googleapis_common_protos-1.72.0-py3-none-any.whl", hash = "sha256:4299c5a82d5ae1a9702ada957347726b167f9f8d1fc352477702a1e851ff4038", size = 297515, upload-time = "2025-11-06T18:29:13.14Z" },
 ]
 
+[package.optional-dependencies]
+grpc = [
+    { name = "grpcio" },
+]
+
 [[package]]
 name = "greenlet"
 version = "3.2.4"
@@ -2245,6 +2386,85 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/e3/a5/6ddab2b4c112be95601c13428db1d8b6608a8b6039816f2ba09c346c08fc/greenlet-3.2.4-cp314-cp314-win_amd64.whl", hash = "sha256:e37ab26028f12dbb0ff65f29a8d3d44a765c61e729647bf2ddfbbed621726f01", size = 303425, upload-time = "2025-08-07T13:32:27.59Z" },
 ]
 
+[[package]]
+name = "grpc-google-iam-v1"
+version = "0.14.4"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "googleapis-common-protos", extra = ["grpc"] },
+    { name = "grpcio" },
+    { name = "protobuf" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/44/4f/d098419ad0bfc06c9ce440575f05aa22d8973b6c276e86ac7890093d3c37/grpc_google_iam_v1-0.14.4.tar.gz", hash = "sha256:392b3796947ed6334e61171d9ab06bf7eb357f554e5fc7556ad7aab6d0e17038", size = 23706, upload-time = "2026-04-01T01:57:49.813Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/89/22/c2dd50c09bf679bd38173656cd4402d2511e563b33bc88f90009cf50613c/grpc_google_iam_v1-0.14.4-py3-none-any.whl", hash = "sha256:412facc320fcbd94034b4df3d557662051d4d8adfa86e0ddb4dca70a3f739964", size = 32675, upload-time = "2026-04-01T01:57:47.69Z" },
+]
+
+[[package]]
+name = "grpcio"
+version = "1.80.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b7/48/af6173dbca4454f4637a4678b67f52ca7e0c1ed7d5894d89d434fecede05/grpcio-1.80.0.tar.gz", hash = "sha256:29aca15edd0688c22ba01d7cc01cb000d72b2033f4a3c72a81a19b56fd143257", size = 12978905, upload-time = "2026-03-30T08:49:10.502Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/5d/db/1d56e5f5823257b291962d6c0ce106146c6447f405b60b234c4f222a7cde/grpcio-1.80.0-cp311-cp311-linux_armv7l.whl", hash = "sha256:dfab85db094068ff42e2a3563f60ab3dddcc9d6488a35abf0132daec13209c8a", size = 6055009, upload-time = "2026-03-30T08:46:46.265Z" },
+    { url = "https://files.pythonhosted.org/packages/6e/18/c83f3cad64c5ca63bca7e91e5e46b0d026afc5af9d0a9972472ceba294b3/grpcio-1.80.0-cp311-cp311-macosx_11_0_universal2.whl", hash = "sha256:5c07e82e822e1161354e32da2662f741a4944ea955f9f580ec8fb409dd6f6060", size = 12035295, upload-time = "2026-03-30T08:46:49.099Z" },
+    { url = "https://files.pythonhosted.org/packages/0f/8e/e14966b435be2dda99fbe89db9525ea436edc79780431a1c2875a3582644/grpcio-1.80.0-cp311-cp311-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:ba0915d51fd4ced2db5ff719f84e270afe0e2d4c45a7bdb1e8d036e4502928c2", size = 6610297, upload-time = "2026-03-30T08:46:52.123Z" },
+    { url = "https://files.pythonhosted.org/packages/cc/26/d5eb38f42ce0e3fdc8174ea4d52036ef8d58cc4426cb800f2610f625dd75/grpcio-1.80.0-cp311-cp311-manylinux2014_i686.manylinux_2_17_i686.whl", hash = "sha256:3cb8130ba457d2aa09fa6b7c3ed6b6e4e6a2685fce63cb803d479576c4d80e21", size = 7300208, upload-time = "2026-03-30T08:46:54.859Z" },
+    { url = "https://files.pythonhosted.org/packages/25/51/bd267c989f85a17a5b3eea65a6feb4ff672af41ca614e5a0279cc0ea381c/grpcio-1.80.0-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:09e5e478b3d14afd23f12e49e8b44c8684ac3c5f08561c43a5b9691c54d136ab", size = 6813442, upload-time = "2026-03-30T08:46:57.056Z" },
+    { url = "https://files.pythonhosted.org/packages/9e/d9/d80eef735b19e9169e30164bbf889b46f9df9127598a83d174eb13a48b26/grpcio-1.80.0-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:00168469238b022500e486c1c33916acf2f2a9b2c022202cf8a1885d2e3073c1", size = 7414743, upload-time = "2026-03-30T08:46:59.682Z" },
+    { url = "https://files.pythonhosted.org/packages/de/f2/567f5bd5054398ed6b0509b9a30900376dcf2786bd936812098808b49d8d/grpcio-1.80.0-cp311-cp311-musllinux_1_2_i686.whl", hash = "sha256:8502122a3cc1714038e39a0b071acb1207ca7844208d5ea0d091317555ee7106", size = 8426046, upload-time = "2026-03-30T08:47:02.474Z" },
+    { url = "https://files.pythonhosted.org/packages/62/29/73ef0141b4732ff5eacd68430ff2512a65c004696997f70476a83e548e7e/grpcio-1.80.0-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:ce1794f4ea6cc3ca29463f42d665c32ba1b964b48958a66497917fe9069f26e6", size = 7851641, upload-time = "2026-03-30T08:47:05.462Z" },
+    { url = "https://files.pythonhosted.org/packages/46/69/abbfa360eb229a8623bab5f5a4f8105e445bd38ce81a89514ba55d281ad0/grpcio-1.80.0-cp311-cp311-win32.whl", hash = "sha256:51b4a7189b0bef2aa30adce3c78f09c83526cf3dddb24c6a96555e3b97340440", size = 4154368, upload-time = "2026-03-30T08:47:08.027Z" },
+    { url = "https://files.pythonhosted.org/packages/6f/d4/ae92206d01183b08613e846076115f5ac5991bae358d2a749fa864da5699/grpcio-1.80.0-cp311-cp311-win_amd64.whl", hash = "sha256:02e64bb0bb2da14d947a49e6f120a75e947250aebe65f9629b62bb1f5c14e6e9", size = 4894235, upload-time = "2026-03-30T08:47:10.839Z" },
+    { url = "https://files.pythonhosted.org/packages/5c/e8/a2b749265eb3415abc94f2e619bbd9e9707bebdda787e61c593004ec927a/grpcio-1.80.0-cp312-cp312-linux_armv7l.whl", hash = "sha256:c624cc9f1008361014378c9d776de7182b11fe8b2e5a81bc69f23a295f2a1ad0", size = 6015616, upload-time = "2026-03-30T08:47:13.428Z" },
+    { url = "https://files.pythonhosted.org/packages/3e/97/b1282161a15d699d1e90c360df18d19165a045ce1c343c7f313f5e8a0b77/grpcio-1.80.0-cp312-cp312-macosx_11_0_universal2.whl", hash = "sha256:f49eddcac43c3bf350c0385366a58f36bed8cc2c0ec35ef7b74b49e56552c0c2", size = 12014204, upload-time = "2026-03-30T08:47:15.873Z" },
+    { url = "https://files.pythonhosted.org/packages/6e/5e/d319c6e997b50c155ac5a8cb12f5173d5b42677510e886d250d50264949d/grpcio-1.80.0-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:d334591df610ab94714048e0d5b4f3dd5ad1bee74dfec11eee344220077a79de", size = 6563866, upload-time = "2026-03-30T08:47:18.588Z" },
+    { url = "https://files.pythonhosted.org/packages/ae/f6/fdd975a2cb4d78eb67769a7b3b3830970bfa2e919f1decf724ae4445f42c/grpcio-1.80.0-cp312-cp312-manylinux2014_i686.manylinux_2_17_i686.whl", hash = "sha256:0cb517eb1d0d0aaf1d87af7cc5b801d686557c1d88b2619f5e31fab3c2315921", size = 7273060, upload-time = "2026-03-30T08:47:21.113Z" },
+    { url = "https://files.pythonhosted.org/packages/db/f0/a3deb5feba60d9538a962913e37bd2e69a195f1c3376a3dd44fe0427e996/grpcio-1.80.0-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:4e78c4ac0d97dc2e569b2f4bcbbb447491167cb358d1a389fc4af71ab6f70411", size = 6782121, upload-time = "2026-03-30T08:47:23.827Z" },
+    { url = "https://files.pythonhosted.org/packages/ca/84/36c6dcfddc093e108141f757c407902a05085e0c328007cb090d56646cdf/grpcio-1.80.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:2ed770b4c06984f3b47eb0517b1c69ad0b84ef3f40128f51448433be904634cd", size = 7383811, upload-time = "2026-03-30T08:47:26.517Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/ef/f3a77e3dc5b471a0ec86c564c98d6adfa3510d38f8ee99010410858d591e/grpcio-1.80.0-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:256507e2f524092f1473071a05e65a5b10d84b82e3ff24c5b571513cfaa61e2f", size = 8393860, upload-time = "2026-03-30T08:47:29.439Z" },
+    { url = "https://files.pythonhosted.org/packages/9b/8d/9d4d27ed7f33d109c50d6b5ce578a9914aa68edab75d65869a17e630a8d1/grpcio-1.80.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:9a6284a5d907c37db53350645567c522be314bac859a64a7a5ca63b77bb7958f", size = 7830132, upload-time = "2026-03-30T08:47:33.254Z" },
+    { url = "https://files.pythonhosted.org/packages/14/e4/9990b41c6d7a44e1e9dee8ac11d7a9802ba1378b40d77468a7761d1ad288/grpcio-1.80.0-cp312-cp312-win32.whl", hash = "sha256:c71309cfce2f22be26aa4a847357c502db6c621f1a49825ae98aa0907595b193", size = 4140904, upload-time = "2026-03-30T08:47:35.319Z" },
+    { url = "https://files.pythonhosted.org/packages/2f/2c/296f6138caca1f4b92a31ace4ae1b87dab692fc16a7a3417af3bb3c805bf/grpcio-1.80.0-cp312-cp312-win_amd64.whl", hash = "sha256:9fe648599c0e37594c4809d81a9e77bd138cc82eb8baa71b6a86af65426723ff", size = 4880944, upload-time = "2026-03-30T08:47:37.831Z" },
+    { url = "https://files.pythonhosted.org/packages/2f/3a/7c3c25789e3f069e581dc342e03613c5b1cb012c4e8c7d9d5cf960a75856/grpcio-1.80.0-cp313-cp313-linux_armv7l.whl", hash = "sha256:e9e408fc016dffd20661f0126c53d8a31c2821b5c13c5d67a0f5ed5de93319ad", size = 6017243, upload-time = "2026-03-30T08:47:40.075Z" },
+    { url = "https://files.pythonhosted.org/packages/04/19/21a9806eb8240e174fd1ab0cd5b9aa948bb0e05c2f2f55f9d5d7405e6d08/grpcio-1.80.0-cp313-cp313-macosx_11_0_universal2.whl", hash = "sha256:92d787312e613754d4d8b9ca6d3297e69994a7912a32fa38c4c4e01c272974b0", size = 12010840, upload-time = "2026-03-30T08:47:43.11Z" },
+    { url = "https://files.pythonhosted.org/packages/18/3a/23347d35f76f639e807fb7a36fad3068aed100996849a33809591f26eca6/grpcio-1.80.0-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:8ac393b58aa16991a2f1144ec578084d544038c12242da3a215966b512904d0f", size = 6567644, upload-time = "2026-03-30T08:47:46.806Z" },
+    { url = "https://files.pythonhosted.org/packages/ff/40/96e07ecb604a6a67ae6ab151e3e35b132875d98bc68ec65f3e5ab3e781d7/grpcio-1.80.0-cp313-cp313-manylinux2014_i686.manylinux_2_17_i686.whl", hash = "sha256:68e5851ac4b9afe07e7f84483803ad167852570d65326b34d54ca560bfa53fb6", size = 7277830, upload-time = "2026-03-30T08:47:49.643Z" },
+    { url = "https://files.pythonhosted.org/packages/9b/e2/da1506ecea1f34a5e365964644b35edef53803052b763ca214ba3870c856/grpcio-1.80.0-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:873ff5d17d68992ef6605330127425d2fc4e77e612fa3c3e0ed4e668685e3140", size = 6783216, upload-time = "2026-03-30T08:47:52.817Z" },
+    { url = "https://files.pythonhosted.org/packages/44/83/3b20ff58d0c3b7f6caaa3af9a4174d4023701df40a3f39f7f1c8e7c48f9d/grpcio-1.80.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:2bea16af2750fd0a899bf1abd9022244418b55d1f37da2202249ba4ba673838d", size = 7385866, upload-time = "2026-03-30T08:47:55.687Z" },
+    { url = "https://files.pythonhosted.org/packages/47/45/55c507599c5520416de5eefecc927d6a0d7af55e91cfffb2e410607e5744/grpcio-1.80.0-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:ba0db34f7e1d803a878284cd70e4c63cb6ae2510ba51937bf8f45ba997cefcf7", size = 8391602, upload-time = "2026-03-30T08:47:58.303Z" },
+    { url = "https://files.pythonhosted.org/packages/10/bb/dd06f4c24c01db9cf11341b547d0a016b2c90ed7dbbb086a5710df7dd1d7/grpcio-1.80.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:8eb613f02d34721f1acf3626dfdb3545bd3c8505b0e52bf8b5710a28d02e8aa7", size = 7826752, upload-time = "2026-03-30T08:48:01.311Z" },
+    { url = "https://files.pythonhosted.org/packages/f9/1e/9d67992ba23371fd63d4527096eb8c6b76d74d52b500df992a3343fd7251/grpcio-1.80.0-cp313-cp313-win32.whl", hash = "sha256:93b6f823810720912fd131f561f91f5fed0fda372b6b7028a2681b8194d5d294", size = 4142310, upload-time = "2026-03-30T08:48:04.594Z" },
+    { url = "https://files.pythonhosted.org/packages/cf/e6/283326a27da9e2c3038bc93eeea36fb118ce0b2d03922a9cda6688f53c5b/grpcio-1.80.0-cp313-cp313-win_amd64.whl", hash = "sha256:e172cf795a3ba5246d3529e4d34c53db70e888fa582a8ffebd2e6e48bc0cba50", size = 4882833, upload-time = "2026-03-30T08:48:07.363Z" },
+    { url = "https://files.pythonhosted.org/packages/c5/6d/e65307ce20f5a09244ba9e9d8476e99fb039de7154f37fb85f26978b59c3/grpcio-1.80.0-cp314-cp314-linux_armv7l.whl", hash = "sha256:3d4147a97c8344d065d01bbf8b6acec2cf86fb0400d40696c8bdad34a64ffc0e", size = 6017376, upload-time = "2026-03-30T08:48:10.005Z" },
+    { url = "https://files.pythonhosted.org/packages/69/10/9cef5d9650c72625a699c549940f0abb3c4bfdb5ed45a5ce431f92f31806/grpcio-1.80.0-cp314-cp314-macosx_11_0_universal2.whl", hash = "sha256:d8e11f167935b3eb089ac9038e1a063e6d7dbe995c0bb4a661e614583352e76f", size = 12018133, upload-time = "2026-03-30T08:48:12.927Z" },
+    { url = "https://files.pythonhosted.org/packages/04/82/983aabaad82ba26113caceeb9091706a0696b25da004fe3defb5b346e15b/grpcio-1.80.0-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:f14b618fc30de822681ee986cfdcc2d9327229dc4c98aed16896761cacd468b9", size = 6574748, upload-time = "2026-03-30T08:48:16.386Z" },
+    { url = "https://files.pythonhosted.org/packages/07/d7/031666ef155aa0bf399ed7e19439656c38bbd143779ae0861b038ce82abd/grpcio-1.80.0-cp314-cp314-manylinux2014_i686.manylinux_2_17_i686.whl", hash = "sha256:4ed39fbdcf9b87370f6e8df4e39ca7b38b3e5e9d1b0013c7b6be9639d6578d14", size = 7277711, upload-time = "2026-03-30T08:48:19.627Z" },
+    { url = "https://files.pythonhosted.org/packages/e8/43/f437a78f7f4f1d311804189e8f11fb311a01049b2e08557c1068d470cb2e/grpcio-1.80.0-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:2dcc70e9f0ba987526e8e8603a610fb4f460e42899e74e7a518bf3c68fe1bf05", size = 6785372, upload-time = "2026-03-30T08:48:22.373Z" },
+    { url = "https://files.pythonhosted.org/packages/93/3d/f6558e9c6296cb4227faa5c43c54a34c68d32654b829f53288313d16a86e/grpcio-1.80.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:448c884b668b868562b1bda833c5fce6272d26e1926ec46747cda05741d302c1", size = 7395268, upload-time = "2026-03-30T08:48:25.638Z" },
+    { url = "https://files.pythonhosted.org/packages/06/21/0fdd77e84720b08843c371a2efa6f2e19dbebf56adc72df73d891f5506f0/grpcio-1.80.0-cp314-cp314-musllinux_1_2_i686.whl", hash = "sha256:a1dc80fe55685b4a543555e6eef975303b36c8db1023b1599b094b92aa77965f", size = 8392000, upload-time = "2026-03-30T08:48:28.974Z" },
+    { url = "https://files.pythonhosted.org/packages/f5/68/67f4947ed55d2e69f2cc199ab9fd85e0a0034d813bbeef84df6d2ba4d4b7/grpcio-1.80.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:31b9ac4ad1aa28ffee5503821fafd09e4da0a261ce1c1281c6c8da0423c83b6e", size = 7828477, upload-time = "2026-03-30T08:48:32.054Z" },
+    { url = "https://files.pythonhosted.org/packages/44/b6/8d4096691b2e385e8271911a0de4f35f0a6c7d05aff7098e296c3de86939/grpcio-1.80.0-cp314-cp314-win32.whl", hash = "sha256:367ce30ba67d05e0592470428f0ec1c31714cab9ef19b8f2e37be1f4c7d32fae", size = 4218563, upload-time = "2026-03-30T08:48:34.538Z" },
+    { url = "https://files.pythonhosted.org/packages/e5/8c/bbe6baf2557262834f2070cf668515fa308b2d38a4bbf771f8f7872a7036/grpcio-1.80.0-cp314-cp314-win_amd64.whl", hash = "sha256:3b01e1f5464c583d2f567b2e46ff0d516ef979978f72091fd81f5ab7fa6e2e7f", size = 5019457, upload-time = "2026-03-30T08:48:37.308Z" },
+]
+
+[[package]]
+name = "grpcio-status"
+version = "1.80.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "googleapis-common-protos" },
+    { name = "grpcio" },
+    { name = "protobuf" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b1/ed/105f619bdd00cb47a49aa2feea6232ea2bbb04199d52a22cc6a7d603b5cb/grpcio_status-1.80.0.tar.gz", hash = "sha256:df73802a4c89a3ea88aa2aff971e886fccce162bc2e6511408b3d67a144381cd", size = 13901, upload-time = "2026-03-30T08:54:34.784Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/76/80/58cd2dfc19a07d022abe44bde7c365627f6c7cb6f692ada6c65ca437d09a/grpcio_status-1.80.0-py3-none-any.whl", hash = "sha256:4b56990363af50dbf2c2ebb80f1967185c07d87aa25aa2bea45ddb75fc181dbe", size = 14638, upload-time = "2026-03-30T08:54:01.569Z" },
+]
+
 [[package]]
 name = "h11"
 version = "0.16.0"
@@ -3141,6 +3361,11 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/e6/05/3516cc7386b220d388aa0bd833308c677e94eceb82b2756dd95e06f6a13f/litellm-1.81.6-py3-none-any.whl", hash = "sha256:573206ba194d49a1691370ba33f781671609ac77c35347f8a0411d852cf6341a", size = 12224343, upload-time = "2026-02-01T04:02:23.704Z" },
 ]
 
+[package.optional-dependencies]
+google = [
+    { name = "google-cloud-aiplatform" },
+]
+
 [[package]]
 name = "locket"
 version = "1.0.0"
@@ -4209,7 +4434,7 @@ dependencies = [
     { name = "fastapi" },
     { name = "google-genai" },
     { name = "kubernetes" },
-    { name = "litellm" },
+    { name = "litellm", extra = ["google"] },
     { name = "openai" },
     { name = "prometheus-client" },
     { name = "prometheus-fastapi-instrumentator" },
@@ -4427,7 +4652,7 @@ requires-dist = [
     { name = "langchain-core", marker = "extra == 'backend'", specifier = "==1.2.11" },
     { name = "langfuse", marker = "extra == 'backend'", specifier = "==3.10.0" },
     { name = "lazy-imports", marker = "extra == 'backend'", specifier = "==1.0.1" },
-    { name = "litellm", specifier = "==1.81.6" },
+    { name = "litellm", extras = ["google"], specifier = "==1.81.6" },
     { name = "lxml", marker = "extra == 'backend'", specifier = "==5.3.0" },
     { name = "mako", marker = "extra == 'backend'", specifier = "==1.2.4" },
     { name = "manygo", marker = "extra == 'dev'", specifier = "==0.2.0" },

web/jest.config.js

@@ -53,6 +53,8 @@ const sharedConfig = {
         // Testing & Mocking
         "msw",
         "until-async",
+        // Language Detection
+        "linguist-languages",
         // Markdown & Syntax Highlighting
         "react-markdown",
         "remark-.*", // All remark packages

web/package-lock.json

@@ -59,8 +59,9 @@
         "lowlight": "^3.3.0",
         "lucide-react": "^0.454.0",
         "mdast-util-find-and-replace": "^3.0.1",
+        "mime": "^4.1.0",
         "motion": "^12.29.0",
-        "next": "16.1.6",
+        "next": "16.1.7",
         "next-themes": "^0.4.4",
         "postcss": "^8.5.6",
         "posthog-js": "^1.176.0",
@@ -2395,9 +2396,9 @@
       }
     },
     "node_modules/@next/env": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/@next/env/-/env-16.1.6.tgz",
-      "integrity": "sha512-N1ySLuZjnAtN3kFnwhAwPvZah8RJxKasD7x1f8shFqhncnWZn4JMfg37diLNuoHsLAlrDfM3g4mawVdtAG8XLQ==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/@next/env/-/env-16.1.7.tgz",
+      "integrity": "sha512-rJJbIdJB/RQr2F1nylZr/PJzamvNNhfr3brdKP6s/GW850jbtR70QlSfFselvIBbcPUOlQwBakexjFzqLzF6pg==",
       "license": "MIT"
     },
     "node_modules/@next/eslint-plugin-next": {
@@ -2441,9 +2442,9 @@
       }
     },
     "node_modules/@next/swc-darwin-arm64": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/@next/swc-darwin-arm64/-/swc-darwin-arm64-16.1.6.tgz",
-      "integrity": "sha512-wTzYulosJr/6nFnqGW7FrG3jfUUlEf8UjGA0/pyypJl42ExdVgC6xJgcXQ+V8QFn6niSG2Pb8+MIG1mZr2vczw==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/@next/swc-darwin-arm64/-/swc-darwin-arm64-16.1.7.tgz",
+      "integrity": "sha512-b2wWIE8sABdyafc4IM8r5Y/dS6kD80JRtOGrUiKTsACFQfWWgUQ2NwoUX1yjFMXVsAwcQeNpnucF2ZrujsBBPg==",
       "cpu": [
         "arm64"
       ],
@@ -2457,9 +2458,9 @@
       }
     },
     "node_modules/@next/swc-darwin-x64": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/@next/swc-darwin-x64/-/swc-darwin-x64-16.1.6.tgz",
-      "integrity": "sha512-BLFPYPDO+MNJsiDWbeVzqvYd4NyuRrEYVB5k2N3JfWncuHAy2IVwMAOlVQDFjj+krkWzhY2apvmekMkfQR0CUQ==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/@next/swc-darwin-x64/-/swc-darwin-x64-16.1.7.tgz",
+      "integrity": "sha512-zcnVaaZulS1WL0Ss38R5Q6D2gz7MtBu8GZLPfK+73D/hp4GFMrC2sudLky1QibfV7h6RJBJs/gOFvYP0X7UVlQ==",
       "cpu": [
         "x64"
       ],
@@ -2473,9 +2474,9 @@
       }
     },
     "node_modules/@next/swc-linux-arm64-gnu": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/@next/swc-linux-arm64-gnu/-/swc-linux-arm64-gnu-16.1.6.tgz",
-      "integrity": "sha512-OJYkCd5pj/QloBvoEcJ2XiMnlJkRv9idWA/j0ugSuA34gMT6f5b7vOiCQHVRpvStoZUknhl6/UxOXL4OwtdaBw==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/@next/swc-linux-arm64-gnu/-/swc-linux-arm64-gnu-16.1.7.tgz",
+      "integrity": "sha512-2ant89Lux/Q3VyC8vNVg7uBaFVP9SwoK2jJOOR0L8TQnX8CAYnh4uctAScy2Hwj2dgjVHqHLORQZJ2wH6VxhSQ==",
       "cpu": [
         "arm64"
       ],
@@ -2489,9 +2490,9 @@
       }
     },
     "node_modules/@next/swc-linux-arm64-musl": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/@next/swc-linux-arm64-musl/-/swc-linux-arm64-musl-16.1.6.tgz",
-      "integrity": "sha512-S4J2v+8tT3NIO9u2q+S0G5KdvNDjXfAv06OhfOzNDaBn5rw84DGXWndOEB7d5/x852A20sW1M56vhC/tRVbccQ==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/@next/swc-linux-arm64-musl/-/swc-linux-arm64-musl-16.1.7.tgz",
+      "integrity": "sha512-uufcze7LYv0FQg9GnNeZ3/whYfo+1Q3HnQpm16o6Uyi0OVzLlk2ZWoY7j07KADZFY8qwDbsmFnMQP3p3+Ftprw==",
       "cpu": [
         "arm64"
       ],
@@ -2505,9 +2506,9 @@
       }
     },
     "node_modules/@next/swc-linux-x64-gnu": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/@next/swc-linux-x64-gnu/-/swc-linux-x64-gnu-16.1.6.tgz",
-      "integrity": "sha512-2eEBDkFlMMNQnkTyPBhQOAyn2qMxyG2eE7GPH2WIDGEpEILcBPI/jdSv4t6xupSP+ot/jkfrCShLAa7+ZUPcJQ==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/@next/swc-linux-x64-gnu/-/swc-linux-x64-gnu-16.1.7.tgz",
+      "integrity": "sha512-KWVf2gxYvHtvuT+c4MBOGxuse5TD7DsMFYSxVxRBnOzok/xryNeQSjXgxSv9QpIVlaGzEn/pIuI6Koosx8CGWA==",
       "cpu": [
         "x64"
       ],
@@ -2521,9 +2522,9 @@
       }
     },
     "node_modules/@next/swc-linux-x64-musl": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/@next/swc-linux-x64-musl/-/swc-linux-x64-musl-16.1.6.tgz",
-      "integrity": "sha512-oicJwRlyOoZXVlxmIMaTq7f8pN9QNbdes0q2FXfRsPhfCi8n8JmOZJm5oo1pwDaFbnnD421rVU409M3evFbIqg==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/@next/swc-linux-x64-musl/-/swc-linux-x64-musl-16.1.7.tgz",
+      "integrity": "sha512-HguhaGwsGr1YAGs68uRKc4aGWxLET+NevJskOcCAwXbwj0fYX0RgZW2gsOCzr9S11CSQPIkxmoSbuVaBp4Z3dA==",
       "cpu": [
         "x64"
       ],
@@ -2537,9 +2538,9 @@
       }
     },
     "node_modules/@next/swc-win32-arm64-msvc": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/@next/swc-win32-arm64-msvc/-/swc-win32-arm64-msvc-16.1.6.tgz",
-      "integrity": "sha512-gQmm8izDTPgs+DCWH22kcDmuUp7NyiJgEl18bcr8irXA5N2m2O+JQIr6f3ct42GOs9c0h8QF3L5SzIxcYAAXXw==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/@next/swc-win32-arm64-msvc/-/swc-win32-arm64-msvc-16.1.7.tgz",
+      "integrity": "sha512-S0n3KrDJokKTeFyM/vGGGR8+pCmXYrjNTk2ZozOL1C/JFdfUIL9O1ATaJOl5r2POe56iRChbsszrjMAdWSv7kQ==",
       "cpu": [
         "arm64"
       ],
@@ -2553,9 +2554,9 @@
       }
     },
     "node_modules/@next/swc-win32-x64-msvc": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/@next/swc-win32-x64-msvc/-/swc-win32-x64-msvc-16.1.6.tgz",
-      "integrity": "sha512-NRfO39AIrzBnixKbjuo2YiYhB6o9d8v/ymU9m/Xk8cyVk+k7XylniXkHwjs4s70wedVffc6bQNbufk5v0xEm0A==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/@next/swc-win32-x64-msvc/-/swc-win32-x64-msvc-16.1.7.tgz",
+      "integrity": "sha512-mwgtg8CNZGYm06LeEd+bNnOUfwOyNem/rOiP14Lsz+AnUY92Zq/LXwtebtUiaeVkhbroRCQ0c8GlR4UT1U+0yg==",
       "cpu": [
         "x64"
       ],
@@ -9225,7 +9226,9 @@
       }
     },
     "node_modules/flatted": {
-      "version": "3.3.3",
+      "version": "3.4.1",
+      "resolved": "https://registry.npmjs.org/flatted/-/flatted-3.4.1.tgz",
+      "integrity": "sha512-IxfVbRFVlV8V/yRaGzk0UVIcsKKHMSfYw66T/u4nTwlWteQePsxe//LjudR1AMX4tZW3WFCh3Zqa/sjlqpbURQ==",
       "dev": true,
       "license": "ISC"
     },
@@ -12746,6 +12749,21 @@
         "url": "https://github.com/sponsors/jonschlinkert"
       }
     },
+    "node_modules/mime": {
+      "version": "4.1.0",
+      "resolved": "https://registry.npmjs.org/mime/-/mime-4.1.0.tgz",
+      "integrity": "sha512-X5ju04+cAzsojXKes0B/S4tcYtFAJ6tTMuSPBEn9CPGlrWr8Fiw7qYeLT0XyH80HSoAoqWCaz+MWKh22P7G1cw==",
+      "funding": [
+        "https://github.com/sponsors/broofa"
+      ],
+      "license": "MIT",
+      "bin": {
+        "mime": "bin/cli.js"
+      },
+      "engines": {
+        "node": ">=16"
+      }
+    },
     "node_modules/mime-db": {
       "version": "1.52.0",
       "license": "MIT",
@@ -12911,14 +12929,14 @@
       "license": "MIT"
     },
     "node_modules/next": {
-      "version": "16.1.6",
-      "resolved": "https://registry.npmjs.org/next/-/next-16.1.6.tgz",
-      "integrity": "sha512-hkyRkcu5x/41KoqnROkfTm2pZVbKxvbZRuNvKXLRXxs3VfyO0WhY50TQS40EuKO9SW3rBj/sF3WbVwDACeMZyw==",
+      "version": "16.1.7",
+      "resolved": "https://registry.npmjs.org/next/-/next-16.1.7.tgz",
+      "integrity": "sha512-WM0L7WrSvKwoLegLYr6V+mz+RIofqQgVAfHhMp9a88ms0cFX8iX9ew+snpWlSBwpkURJOUdvCEt3uLl3NNzvWg==",
       "license": "MIT",
       "dependencies": {
-        "@next/env": "16.1.6",
+        "@next/env": "16.1.7",
         "@swc/helpers": "0.5.15",
-        "baseline-browser-mapping": "^2.8.3",
+        "baseline-browser-mapping": "^2.9.19",
         "caniuse-lite": "^1.0.30001579",
         "postcss": "8.4.31",
         "styled-jsx": "5.1.6"
@@ -12930,14 +12948,14 @@
         "node": ">=20.9.0"
       },
       "optionalDependencies": {
-        "@next/swc-darwin-arm64": "16.1.6",
-        "@next/swc-darwin-x64": "16.1.6",
-        "@next/swc-linux-arm64-gnu": "16.1.6",
-        "@next/swc-linux-arm64-musl": "16.1.6",
-        "@next/swc-linux-x64-gnu": "16.1.6",
-        "@next/swc-linux-x64-musl": "16.1.6",
-        "@next/swc-win32-arm64-msvc": "16.1.6",
-        "@next/swc-win32-x64-msvc": "16.1.6",
+        "@next/swc-darwin-arm64": "16.1.7",
+        "@next/swc-darwin-x64": "16.1.7",
+        "@next/swc-linux-arm64-gnu": "16.1.7",
+        "@next/swc-linux-arm64-musl": "16.1.7",
+        "@next/swc-linux-x64-gnu": "16.1.7",
+        "@next/swc-linux-x64-musl": "16.1.7",
+        "@next/swc-win32-arm64-msvc": "16.1.7",
+        "@next/swc-win32-x64-msvc": "16.1.7",
         "sharp": "^0.34.4"
       },
       "peerDependencies": {

web/package.json

@@ -75,8 +75,9 @@
     "lowlight": "^3.3.0",
     "lucide-react": "^0.454.0",
     "mdast-util-find-and-replace": "^3.0.1",
+    "mime": "^4.1.0",
     "motion": "^12.29.0",
-    "next": "16.1.6",
+    "next": "16.1.7",
     "next-themes": "^0.4.4",
     "postcss": "^8.5.6",
     "posthog-js": "^1.176.0",

web/src/app/admin/connector/[ccPairId]/page.tsx

@@ -626,10 +626,7 @@ function Main({ ccPairId }: { ccPairId: number }) {
           <div className="w-[200px]">
             <div className="text-sm font-medium mb-1">Last Indexed</div>
             <div className="text-sm text-text-default">
-              {timeAgo(
-                indexAttempts?.find((attempt) => attempt.status === "success")
-                  ?.time_started
-              ) ?? "-"}
+              {timeAgo(ccPair?.last_indexed) ?? "-"}
             </div>
           </div>
 

web/src/app/admin/connectors/[connector]/pages/utils/google_site.ts

@@ -21,10 +21,13 @@ export const submitGoogleSite = async (
       formData.append("files", file);
     });
 
-    const response = await fetch("/api/manage/admin/connector/file/upload", {
-      method: "POST",
-      body: formData,
-    });
+    const response = await fetch(
+      "/api/manage/admin/connector/file/upload?unzip=false",
+      {
+        method: "POST",
+        body: formData,
+      }
+    );
     const responseJson = await response.json();
     if (!response.ok) {
       toast.error(`Unable to upload files - ${responseJson.detail}`);

web/src/app/app/message/FileDisplay.tsx

@@ -6,7 +6,7 @@ import { ChatFileType, FileDescriptor } from "@/app/app/interfaces";
 import Attachment from "@/refresh-components/Attachment";
 import { InMessageImage } from "@/app/app/components/files/images/InMessageImage";
 import CsvContent from "@/components/tools/CSVContent";
-import TextViewModal from "@/sections/modals/TextViewModal";
+import PreviewModal from "@/sections/modals/PreviewModal";
 import { MinimalOnyxDocument } from "@/lib/search/interfaces";
 import ExpandableContentWrapper from "@/components/tools/ExpandableContentWrapper";
 
@@ -52,7 +52,7 @@ export default function FileDisplay({ files }: FileDisplayProps) {
   return (
     <>
       {previewingFile && (
-        <TextViewModal
+        <PreviewModal
           presentingDocument={presentingDocument}
           onClose={() => setPreviewingFile(null)}
         />

web/src/app/app/message/custom-code-styles.css

@@ -1,7 +1,7 @@
 /* Light mode syntax highlighting (Atom One Light) */
 .hljs {
   color: #383a42 !important;
-  background: #fafafa !important;
+  background: var(--background-code-01) !important;
 }
 
 .hljs-comment,
@@ -77,7 +77,7 @@
 /* Dark mode syntax highlighting (Atom One Dark) */
 .dark .hljs {
   color: #e2e6eb !important;
-  background: #151617 !important;
+  background: var(--background-code-01) !important;
 }
 
 .dark .hljs-comment,

web/src/app/app/shared/[chatId]/SharedChatDisplay.tsx

@@ -11,7 +11,7 @@ import { Callout } from "@/components/ui/callout";
 import OnyxInitializingLoader from "@/components/OnyxInitializingLoader";
 import { Persona } from "@/app/admin/agents/interfaces";
 import { MinimalOnyxDocument } from "@/lib/search/interfaces";
-import TextViewModal from "@/sections/modals/TextViewModal";
+import PreviewModal from "@/sections/modals/PreviewModal";
 import { UNNAMED_CHAT } from "@/lib/constants";
 import Text from "@/refresh-components/texts/Text";
 import useOnMount from "@/hooks/useOnMount";
@@ -64,7 +64,7 @@ export default function SharedChatDisplay({
   return (
     <>
       {presentingDocument && (
-        <TextViewModal
+        <PreviewModal
           presentingDocument={presentingDocument}
           onClose={() => setPresentingDocument(null)}
         />

web/src/app/craft/components/InputBar.tsx

@@ -13,6 +13,7 @@ import {
   type KeyboardEvent,
 } from "react";
 import { useRouter } from "next/navigation";
+import { getPastedFilesIfNoText } from "@/lib/clipboard";
 import { cn, isImageFile } from "@/lib/utils";
 import { Disabled } from "@opal/core";
 import {
@@ -230,21 +231,11 @@ const InputBar = memo(
 
       const handlePaste = useCallback(
         (event: ClipboardEvent) => {
-          const items = event.clipboardData?.items;
-          if (items) {
-            const pastedFiles: File[] = [];
-            for (let i = 0; i < items.length; i++) {
-              const item = items[i];
-              if (item && item.kind === "file") {
-                const file = item.getAsFile();
-                if (file) pastedFiles.push(file);
-              }
-            }
-            if (pastedFiles.length > 0) {
-              event.preventDefault();
-              // Context handles session binding internally
-              uploadFiles(pastedFiles);
-            }
+          const pastedFiles = getPastedFilesIfNoText(event.clipboardData);
+          if (pastedFiles.length > 0) {
+            event.preventDefault();
+            // Context handles session binding internally
+            uploadFiles(pastedFiles);
           }
         },
         [uploadFiles]

web/src/app/craft/onboarding/components/OnboardingLlmSetup.tsx

@@ -34,7 +34,8 @@ export const PROVIDERS: ProviderConfig[] = [
     providerName: LLMProviderName.ANTHROPIC,
     recommended: true,
     models: [
-      { name: "claude-opus-4-6", label: "Claude Opus 4.6", recommended: true },
+      { name: "claude-opus-4-7", label: "Claude Opus 4.7", recommended: true },
+      { name: "claude-opus-4-6", label: "Claude Opus 4.6" },
       { name: "claude-sonnet-4-6", label: "Claude Sonnet 4.6" },
     ],
     apiKeyPlaceholder: "sk-ant-...",

web/src/app/craft/onboarding/constants.ts

@@ -5,12 +5,12 @@
 export interface BuildLlmSelection {
   providerName: string; // e.g., "build-mode-anthropic" (LLMProviderDescriptor.name)
   provider: string; // e.g., "anthropic"
-  modelName: string; // e.g., "claude-opus-4-6"
+  modelName: string; // e.g., "claude-opus-4-7"
 }
 
 // Priority order for smart default LLM selection
 const LLM_SELECTION_PRIORITY = [
-  { provider: "anthropic", modelName: "claude-opus-4-6" },
+  { provider: "anthropic", modelName: "claude-opus-4-7" },
   { provider: "openai", modelName: "gpt-5.2" },
   { provider: "openrouter", modelName: "minimax/minimax-m2.1" },
 ] as const;
@@ -63,10 +63,11 @@ export function getDefaultLlmSelection(
 export const RECOMMENDED_BUILD_MODELS = {
   preferred: {
     provider: "anthropic",
-    modelName: "claude-opus-4-6",
-    displayName: "Claude Opus 4.6",
+    modelName: "claude-opus-4-7",
+    displayName: "Claude Opus 4.7",
   },
   alternatives: [
+    { provider: "anthropic", modelName: "claude-opus-4-6" },
     { provider: "anthropic", modelName: "claude-sonnet-4-6" },
     { provider: "openai", modelName: "gpt-5.2" },
     { provider: "openai", modelName: "gpt-5.1-codex" },
@@ -148,7 +149,8 @@ export const BUILD_MODE_PROVIDERS: BuildModeProvider[] = [
     providerName: "anthropic",
     recommended: true,
     models: [
-      { name: "claude-opus-4-6", label: "Claude Opus 4.6", recommended: true },
+      { name: "claude-opus-4-7", label: "Claude Opus 4.7", recommended: true },
+      { name: "claude-opus-4-6", label: "Claude Opus 4.6" },
       { name: "claude-sonnet-4-6", label: "Claude Sonnet 4.6" },
     ],
     apiKeyPlaceholder: "sk-ant-...",

web/src/app/craft/v1/configure/components/UserLibraryModal.tsx

@@ -272,6 +272,22 @@ export default function UserLibraryModal({
                   </Disabled>
                 </Section>
 
+                {/* The exact cap is controlled by the backend env var
+                    MAX_EMBEDDED_IMAGES_PER_FILE (default 500). This copy is
+                    deliberately vague so it doesn't drift if the limit is
+                    tuned per-deployment; the precise number is surfaced in
+                    the rejection error the server returns. */}
+                <Section
+                  flexDirection="row"
+                  justifyContent="end"
+                  padding={0.5}
+                  height="fit"
+                >
+                  <Text secondaryBody text03>
+                    PDFs with many embedded images may be rejected.
+                  </Text>
+                </Section>
+
                 {isLoading ? (
                   <Section padding={2} height="fit">
                     <Text secondaryBody text03>

web/src/app/css/colors.css

@@ -438,6 +438,9 @@
   --action-text-link-05: var(--blue-50);
   --action-text-danger-05: var(--red-50);
 
+  /* Background / Code */
+  --background-code-01: var(--grey-02);
+
   /* Code */
   --code-code: var(--alpha-grey-100-85);
   --code-comment: var(--alpha-grey-100-35);
@@ -639,6 +642,9 @@
   --action-text-link-05: var(--blue-45);
   --action-text-danger-05: var(--red-45);
 
+  /* Background / Code */
+  --background-code-01: #151617;
+
   /* Code */
   --code-code: var(--alpha-grey-00-85);
   --code-comment: var(--alpha-grey-00-45);

web/src/app/nrf/NRFPage.tsx

@@ -40,7 +40,7 @@ import { SvgUser, SvgMenu, SvgAlertTriangle } from "@opal/icons";
 import { useAppBackground } from "@/providers/AppBackgroundProvider";
 import { MinimalOnyxDocument } from "@/lib/search/interfaces";
 import DocumentsSidebar from "@/sections/document-sidebar/DocumentsSidebar";
-import TextViewModal from "@/sections/modals/TextViewModal";
+import PreviewModal from "@/sections/modals/PreviewModal";
 import { personaIncludesRetrieval } from "@/app/app/services/lib";
 import { useQueryController } from "@/providers/QueryControllerProvider";
 import { eeGated } from "@/ce";
@@ -295,7 +295,6 @@ export default function NRFPage({ isSidePanel = false }: NRFPageProps) {
       // If we already have messages (chat session started), always use chat mode
       // (matches AppPage behavior where existing sessions bypass classification)
       if (hasMessages) {
-        resetInputBar();
         onSubmit({
           message: submittedMessage,
           currentMessageFiles: currentMessageFiles,
@@ -307,7 +306,6 @@ export default function NRFPage({ isSidePanel = false }: NRFPageProps) {
 
       // Build an onChat closure that captures additionalContext for this submission
       const onChat = (chatMessage: string) => {
-        resetInputBar();
         onSubmit({
           message: chatMessage,
           currentMessageFiles: currentMessageFiles,
@@ -326,7 +324,6 @@ export default function NRFPage({ isSidePanel = false }: NRFPageProps) {
       onSubmit,
       currentMessageFiles,
       deepResearchEnabled,
-      resetInputBar,
       submitQuery,
       tabReadingEnabled,
       currentTabUrl,
@@ -537,7 +534,7 @@ export default function NRFPage({ isSidePanel = false }: NRFPageProps) {
 
       {/* Text/document preview modal */}
       {presentingDocument && (
-        <TextViewModal
+        <PreviewModal
           presentingDocument={presentingDocument}
           onClose={() => setPresentingDocument(null)}
         />

web/src/components/chat/MinimalMarkdown.tsx

@@ -4,21 +4,21 @@ import {
   MemoizedLink,
   MemoizedParagraph,
 } from "@/app/app/message/MemoizedTextComponents";
-import React, { useMemo, CSSProperties } from "react";
+import { useMemo, CSSProperties } from "react";
 import ReactMarkdown, { type Components } from "react-markdown";
 import remarkGfm from "remark-gfm";
 import rehypeHighlight from "rehype-highlight";
 import remarkMath from "remark-math";
 import rehypeKatex from "rehype-katex";
 import "katex/dist/katex.min.css";
-import { transformLinkUri } from "@/lib/utils";
+import { cn, transformLinkUri } from "@/lib/utils";
 
 type MinimalMarkdownComponentOverrides = Partial<Components>;
 
 interface MinimalMarkdownProps {
   content: string;
   className?: string;
-  style?: CSSProperties;
+  showHeader?: boolean;
   /**
    * Override specific markdown renderers.
    * Any renderer not provided will fall back to this component's defaults.
@@ -29,7 +29,7 @@ interface MinimalMarkdownProps {
 export default function MinimalMarkdown({
   content,
   className = "",
-  style,
+  showHeader = true,
   components,
 }: MinimalMarkdownProps) {
   const markdownComponents = useMemo(() => {
@@ -43,7 +43,11 @@ export default function MinimalMarkdown({
       code: ({ node, inline, className, children, ...props }: any) => {
         const codeText = extractCodeText(node, content, children);
         return (
-          <CodeBlock className={className} codeText={codeText}>
+          <CodeBlock
+            className={className}
+            codeText={codeText}
+            showHeader={showHeader}
+          >
             {children}
           </CodeBlock>
         );
@@ -54,22 +58,20 @@ export default function MinimalMarkdown({
       ...defaults,
       ...(components ?? {}),
     } satisfies Components;
-  }, [content, components]);
+  }, [content, components, showHeader]);
 
   return (
-    <div style={style || {}} className={`${className}`}>
-      <ReactMarkdown
-        className="prose dark:prose-invert max-w-full text-sm break-words"
-        components={markdownComponents}
-        rehypePlugins={[rehypeHighlight, rehypeKatex]}
-        remarkPlugins={[
-          remarkGfm,
-          [remarkMath, { singleDollarTextMath: false }],
-        ]}
-        urlTransform={transformLinkUri}
-      >
-        {content}
-      </ReactMarkdown>
-    </div>
+    <ReactMarkdown
+      className={cn(
+        "prose dark:prose-invert max-w-full text-sm break-words",
+        className
+      )}
+      components={markdownComponents}
+      rehypePlugins={[rehypeHighlight, rehypeKatex]}
+      remarkPlugins={[remarkGfm, [remarkMath, { singleDollarTextMath: false }]]}
+      urlTransform={transformLinkUri}
+    >
+      {content}
+    </ReactMarkdown>
   );
 }

web/src/components/dateRangeSelectors/AdminDateRangeSelector.tsx

@@ -52,6 +52,7 @@ export const AdminDateRangeSelector = memo(function AdminDateRangeSelector({
         <Popover.Trigger asChild>
           {/* TODO(@raunakab): migrate to opal Button once className/iconClassName is resolved */}
           <Button
+            data-testid="admin-date-range-selector-button"
             secondary
             className={cn("justify-start", !value && "text-muted-foreground")}
             leftIcon={SvgCalendar}

web/src/components/tools/ExpandableContentWrapper.tsx

@@ -6,7 +6,7 @@ import { Button } from "@opal/components";
 import Text from "@/refresh-components/texts/Text";
 import { FileDescriptor } from "@/app/app/interfaces";
 import { cn } from "@/lib/utils";
-import TextViewModal from "@/sections/modals/TextViewModal";
+import PreviewModal from "@/sections/modals/PreviewModal";
 import { MinimalOnyxDocument } from "@/lib/search/interfaces";
 
 export interface ExpandableContentWrapperProps {
@@ -95,7 +95,7 @@ export default function ExpandableContentWrapper({
   return (
     <>
       {expanded && (
-        <TextViewModal
+        <PreviewModal
           presentingDocument={presentingDocument}
           onClose={() => setExpanded(false)}
         />

web/src/lib/clipboard.test.ts

@@ -0,0 +1,89 @@
+import { getPastedFilesIfNoText } from "./clipboard";
+
+type MockClipboardData = Parameters<typeof getPastedFilesIfNoText>[0];
+
+function makeClipboardData({
+  textPlain = "",
+  text = "",
+  files = [],
+}: {
+  textPlain?: string;
+  text?: string;
+  files?: File[];
+}): MockClipboardData {
+  return {
+    items: files.map((file) => ({
+      kind: "file",
+      getAsFile: () => file,
+    })),
+    getData: (format: string) => {
+      if (format === "text/plain") {
+        return textPlain;
+      }
+
+      if (format === "text") {
+        return text;
+      }
+
+      return "";
+    },
+  };
+}
+
+describe("getPastedFilesIfNoText", () => {
+  it("prefers plain text over pasted files when both are present", () => {
+    const imageFile = new File(["slide preview"], "slide.png", {
+      type: "image/png",
+    });
+
+    expect(
+      getPastedFilesIfNoText(
+        makeClipboardData({
+          textPlain: "Welcome to PowerPoint for Mac",
+          files: [imageFile],
+        })
+      )
+    ).toEqual([]);
+  });
+
+  it("falls back to text data when text/plain is empty", () => {
+    const imageFile = new File(["slide preview"], "slide.png", {
+      type: "image/png",
+    });
+
+    expect(
+      getPastedFilesIfNoText(
+        makeClipboardData({
+          text: "Welcome to PowerPoint for Mac",
+          files: [imageFile],
+        })
+      )
+    ).toEqual([]);
+  });
+
+  it("still returns files for image-only pastes", () => {
+    const imageFile = new File(["slide preview"], "slide.png", {
+      type: "image/png",
+    });
+
+    expect(
+      getPastedFilesIfNoText(makeClipboardData({ files: [imageFile] }))
+    ).toEqual([imageFile]);
+  });
+
+  it("ignores whitespace-only text and keeps file pastes working", () => {
+    const imageFile = new File(["slide preview"], "slide.png", {
+      type: "image/png",
+    });
+
+    expect(
+      getPastedFilesIfNoText(
+        makeClipboardData({
+          textPlain: "   ",
+          text: "\n",
+          files: [imageFile],
+        })
+      )
+    ).toEqual([imageFile]);
+  });
+});

web/src/lib/clipboard.ts

@@ -0,0 +1,52 @@
+type ClipboardFileItem = {
+  kind: string;
+  getAsFile: () => File | null;
+};
+
+type ClipboardDataLike = {
+  items?: ArrayLike<ClipboardFileItem> | null;
+  getData: (format: string) => string;
+};
+
+function getClipboardText(
+  clipboardData: ClipboardDataLike,
+  format: "text/plain" | "text"
+): string {
+  try {
+    return clipboardData.getData(format);
+  } catch {
+    return "";
+  }
+}
+
+export function getPastedFilesIfNoText(
+  clipboardData?: ClipboardDataLike | null
+): File[] {
+  if (!clipboardData) {
+    return [];
+  }
+
+  const plainText = getClipboardText(clipboardData, "text/plain").trim();
+  const fallbackText = getClipboardText(clipboardData, "text").trim();
+
+  // Apps like PowerPoint on macOS can place both rendered image data and the
+  // original text on the clipboard. Prefer letting the textarea consume text.
+  if (plainText || fallbackText || !clipboardData.items) {
+    return [];
+  }
+
+  const pastedFiles: File[] = [];
+  for (let i = 0; i < clipboardData.items.length; i++) {
+    const item = clipboardData.items[i];
+    if (item?.kind !== "file") {
+      continue;
+    }
+
+    const file = item.getAsFile();
+    if (file) {
+      pastedFiles.push(file);
+    }
+  }
+
+  return pastedFiles;
+}

web/src/lib/languages.test.ts

@@ -0,0 +1,102 @@
+import {
+  getCodeLanguage,
+  getDataLanguage,
+  getLanguageByMime,
+  isMarkdownFile,
+} from "./languages";
+
+describe("getCodeLanguage", () => {
+  it.each([
+    ["app.py", "python"],
+    ["index.ts", "typescript"],
+    ["main.go", "go"],
+    ["style.css", "css"],
+    ["page.html", "html"],
+    ["App.vue", "vue"],
+    ["lib.rs", "rust"],
+    ["main.cpp", "c++"],
+    ["util.c", "c"],
+    ["script.js", "javascript"],
+  ])("%s → %s", (filename, expected) => {
+    expect(getCodeLanguage(filename)).toBe(expected);
+  });
+
+  it.each([
+    [".h", "c"],
+    [".inc", "php"],
+    [".m", "objective-c"],
+    [".re", "reason"],
+  ])("override: %s → %s", (ext, expected) => {
+    expect(getCodeLanguage(`file${ext}`)).toBe(expected);
+  });
+
+  it("resolves by exact filename when there is no extension", () => {
+    expect(getCodeLanguage("Dockerfile")).toBe("dockerfile");
+    expect(getCodeLanguage("Makefile")).toBe("makefile");
+  });
+
+  it("is case-insensitive for filenames", () => {
+    expect(getCodeLanguage("INDEX.JS")).toBe("javascript");
+    expect(getCodeLanguage("dockerfile")).toBe("dockerfile");
+  });
+
+  it("returns null for unknown extensions", () => {
+    expect(getCodeLanguage("file.xyz123")).toBeNull();
+  });
+
+  it("excludes markdown extensions", () => {
+    expect(getCodeLanguage("README.md")).toBeNull();
+    expect(getCodeLanguage("notes.markdown")).toBeNull();
+  });
+});
+
+describe("getDataLanguage", () => {
+  it.each([
+    ["config.json", "json"],
+    ["config.yaml", "yaml"],
+    ["config.yml", "yaml"],
+    ["config.toml", "toml"],
+    ["data.xml", "xml"],
+    ["data.csv", "csv"],
+  ])("%s → %s", (filename, expected) => {
+    expect(getDataLanguage(filename)).toBe(expected);
+  });
+
+  it("returns null for code files", () => {
+    expect(getDataLanguage("app.py")).toBeNull();
+    expect(getDataLanguage("header.h")).toBeNull();
+    expect(getDataLanguage("view.m")).toBeNull();
+    expect(getDataLanguage("component.re")).toBeNull();
+  });
+});
+
+describe("isMarkdownFile", () => {
+  it("recognises markdown extensions", () => {
+    expect(isMarkdownFile("README.md")).toBe(true);
+    expect(isMarkdownFile("doc.markdown")).toBe(true);
+  });
+
+  it("is case-insensitive", () => {
+    expect(isMarkdownFile("NOTES.MD")).toBe(true);
+  });
+
+  it("rejects non-markdown files", () => {
+    expect(isMarkdownFile("app.py")).toBe(false);
+    expect(isMarkdownFile("data.json")).toBe(false);
+  });
+});
+
+describe("getLanguageByMime", () => {
+  it("resolves known MIME types", () => {
+    expect(getLanguageByMime("text/x-python")).toBe("python");
+    expect(getLanguageByMime("text/javascript")).toBe("javascript");
+  });
+
+  it("strips parameters before matching", () => {
+    expect(getLanguageByMime("text/x-python; charset=utf-8")).toBe("python");
+  });
+
+  it("returns null for unknown MIME types", () => {
+    expect(getLanguageByMime("application/x-unknown-thing")).toBeNull();
+  });
+});

web/src/lib/languages.ts

@@ -7,6 +7,7 @@ interface LinguistLanguage {
   type: string;
   extensions?: string[];
   filenames?: string[];
+  codemirrorMimeType?: string;
 }
 
 interface LanguageMaps {
@@ -14,7 +15,23 @@ interface LanguageMaps {
   filenames: Map<string, string>;
 }
 
-const allLanguages = Object.values(languages) as LinguistLanguage[];
+// Explicit winners for extensions claimed by multiple linguist-languages entries
+// where the "most extensions" heuristic below picks the wrong language.
+const EXTENSION_OVERRIDES: Record<string, string> = {
+  ".h": "c",
+  ".inc": "php",
+  ".m": "objective-c",
+  ".re": "reason",
+  ".rs": "rust",
+};
+
+// Sort so that languages with more extensions (i.e. more general-purpose) win
+// when multiple languages claim the same extension (e.g. Ecmarkup vs HTML both
+// claim .html — HTML should win because it's the canonical language for that
+// extension). Known mis-rankings are patched by EXTENSION_OVERRIDES above.
+const allLanguages = (Object.values(languages) as LinguistLanguage[]).sort(
+  (a, b) => (b.extensions?.length ?? 0) - (a.extensions?.length ?? 0)
+);
 
 // Collect extensions that linguist-languages assigns to "Markdown" so we can
 // exclude them from the code-language map
@@ -25,14 +42,22 @@ const markdownExtensions = new Set(
 );
 
 function buildLanguageMaps(
-  type: string,
+  types: string[],
   excludedExtensions?: Set<string>
 ): LanguageMaps {
+  const typeSet = new Set(types);
   const extensions = new Map<string, string>();
   const filenames = new Map<string, string>();
 
+  if (typeSet.has("programming") || typeSet.has("markup")) {
+    for (const [ext, lang] of Object.entries(EXTENSION_OVERRIDES)) {
+      if (excludedExtensions?.has(ext.toLowerCase())) continue;
+      extensions.set(ext, lang);
+    }
+  }
+
   for (const lang of allLanguages) {
-    if (lang.type !== type) continue;
+    if (!typeSet.has(lang.type)) continue;
 
     const name = lang.name.toLowerCase();
     for (const ext of lang.extensions ?? []) {
@@ -57,13 +82,17 @@ function lookupLanguage(name: string, maps: LanguageMaps): string | null {
   return (ext && maps.extensions.get(ext)) ?? maps.filenames.get(lower) ?? null;
 }
 
-const codeMaps = buildLanguageMaps("programming", markdownExtensions);
-const dataMaps = buildLanguageMaps("data");
+const codeMaps = buildLanguageMaps(
+  ["programming", "markup"],
+  markdownExtensions
+);
+const dataMaps = buildLanguageMaps(["data"]);
 
 /**
  * Returns the language name for a given file name, or null if it's not a
- * recognised code file. Looks up by extension first, then by exact filename
- * (e.g. "Dockerfile", "Makefile"). Runs in O(1).
+ * recognised code or markup file (programming + markup types from
+ * linguist-languages, e.g. Python, HTML, CSS, Vue). Looks up by extension
+ * first, then by exact filename (e.g. "Dockerfile", "Makefile"). Runs in O(1).
  */
 export function getCodeLanguage(name: string): string | null {
   return lookupLanguage(name, codeMaps);
@@ -86,3 +115,20 @@ export function isMarkdownFile(name: string): boolean {
   const ext = name.toLowerCase().match(LANGUAGE_EXT_PATTERN)?.[0];
   return !!ext && markdownExtensions.has(ext);
 }
+
+const mimeToLanguage = new Map<string, string>();
+for (const lang of allLanguages) {
+  if (lang.codemirrorMimeType && !mimeToLanguage.has(lang.codemirrorMimeType)) {
+    mimeToLanguage.set(lang.codemirrorMimeType, lang.name.toLowerCase());
+  }
+}
+
+/**
+ * Returns the language name for a given MIME type using the codemirrorMimeType
+ * field from linguist-languages (~297 entries). Returns null if unrecognised.
+ */
+export function getLanguageByMime(mime: string): string | null {
+  const base = mime.split(";")[0];
+  if (!base) return null;
+  return mimeToLanguage.get(base.trim().toLowerCase()) ?? null;
+}

web/src/refresh-components/ScrollIndicatorDiv.tsx

@@ -6,10 +6,42 @@ import { cn } from "@/lib/utils";
 // Throttle interval for scroll events (~60fps)
 const SCROLL_THROTTLE_MS = 16;
 
+/**
+ * A scrollable container that shows gradient or shadow indicators when
+ * content overflows above or below the visible area.
+ *
+ * HEIGHT CONSTRAINT REQUIREMENT
+ *
+ * This component relies on its inner scroll container having a smaller
+ * clientHeight than its scrollHeight. For that to happen, the entire
+ * ancestor chain must constrain height via flex sizing (flex-1 min-h-0),
+ * NOT via percentage heights (h-full).
+ *
+ * height: 100% resolves to "auto" when the containing block's height is
+ * determined by flex layout (flex-auto, flex-1) rather than an explicit
+ * height property — this is per the CSS spec. When that happens, the
+ * container grows to fit its content and scrollHeight === clientHeight,
+ * making scroll indicators invisible.
+ *
+ * Correct pattern: every ancestor up to the nearest fixed-height boundary
+ * must form an unbroken flex column chain using "flex-1 min-h-0":
+ *
+ *   fixed-height-ancestor  (e.g. h-[500px])
+ *     flex flex-col flex-1 min-h-0   <-- use flex-1, NOT h-full
+ *       ScrollIndicatorDiv
+ *         ...tall content...
+ *
+ * Common mistakes:
+ *  - Using h-full instead of flex-1 min-h-0 anywhere in the chain.
+ *  - Placing this inside a parent with overflow-y: auto (e.g. Modal.Body),
+ *    which becomes the scroll container instead of this component's inner div.
+ */
 export interface ScrollIndicatorDivProps
   extends React.HTMLAttributes<HTMLDivElement> {
   // Mask/Shadow options
   disableIndicators?: boolean;
+  disableTopIndicator?: boolean;
+  disableBottomIndicator?: boolean;
   backgroundColor?: string;
   indicatorHeight?: string;
 
@@ -22,6 +54,8 @@ export interface ScrollIndicatorDivProps
 
 export default function ScrollIndicatorDiv({
   disableIndicators = false,
+  disableTopIndicator = false,
+  disableBottomIndicator = false,
   backgroundColor = "var(--background-tint-02)",
   indicatorHeight = "3rem",
   variant = "gradient",
@@ -77,13 +111,19 @@ export default function ScrollIndicatorDiv({
     // Update on scroll (throttled)
     container.addEventListener("scroll", handleScroll, { passive: true });
 
-    // Update on resize (in case content changes)
+    // Update when the container itself resizes
     const resizeObserver = new ResizeObserver(updateScrollIndicators);
     resizeObserver.observe(container);
 
+    // Update when descendants change (e.g. syntax highlighting mutates the
+    // DOM after initial render, which changes scrollHeight without firing
+    // resize or scroll events on the container).
+    const mutationObserver = new MutationObserver(handleScroll);
+
     return () => {
       container.removeEventListener("scroll", handleScroll);
       resizeObserver.disconnect();
+      mutationObserver.disconnect();
       if (throttleTimeoutRef.current) {
         clearTimeout(throttleTimeoutRef.current);
       }
@@ -120,7 +160,7 @@ export default function ScrollIndicatorDiv({
   return (
     <div className="relative flex-1 min-h-0 overflow-y-hidden flex flex-col w-full">
       {/* Top indicator */}
-      {!disableIndicators && showTopIndicator && (
+      {!disableIndicators && !disableTopIndicator && showTopIndicator && (
         <div
           className="absolute top-0 left-0 right-0 z-[20] pointer-events-none transition-opacity duration-200"
           style={getIndicatorStyle("top")}
@@ -141,7 +181,7 @@ export default function ScrollIndicatorDiv({
       </div>
 
       {/* Bottom indicator */}
-      {!disableIndicators && showBottomIndicator && (
+      {!disableIndicators && !disableBottomIndicator && showBottomIndicator && (
         <div
           className="absolute bottom-0 left-0 right-0 z-[20] pointer-events-none transition-opacity duration-200"
           style={getIndicatorStyle("bottom")}

web/src/refresh-components/modals/MemoriesModal.tsx

@@ -112,9 +112,11 @@ function MemoryItem({
             />
           </Disabled>
         </Section>
-        {isFocused && (
+        <div
+          className={isFocused ? "visible" : "invisible h-0 overflow-hidden"}
+        >
           <CharacterCount value={memory.content} limit={MAX_MEMORY_LENGTH} />
-        )}
+        </div>
       </Section>
     </div>
   );

web/src/refresh-pages/AgentEditorPage.tsx

@@ -661,7 +661,7 @@ export default function AgentEditorPage({
     // Sharing
     shared_user_ids: existingAgent?.users?.map((user) => user.id) ?? [],
     shared_group_ids: existingAgent?.groups ?? [],
-    is_public: existingAgent?.is_public ?? true,
+    is_public: existingAgent?.is_public ?? false,
     label_ids: existingAgent?.labels?.map((l) => l.id) ?? [],
     featured: existingAgent?.featured ?? false,
   };
@@ -967,6 +967,14 @@ export default function AgentEditorPage({
           validateOnChange
           validateOnBlur
           validateOnMount
+          initialTouched={{
+            description:
+              initialValues.description.length >
+              MAX_CHARACTERS_AGENT_DESCRIPTION,
+            starter_messages: initialValues.starter_messages.map(
+              (msg) => msg.length > MAX_CHARACTERS_STARTER_MESSAGE
+            ) as unknown as boolean,
+          }}
           initialStatus={{ warnings: {} }}
         >
           {({ isSubmitting, isValid, dirty, values, setFieldValue }) => {
@@ -1201,18 +1209,33 @@ export default function AgentEditorPage({
                           >
                             Cancel
                           </OpalButton>
-                          <Disabled
-                            disabled={
-                              isSubmitting ||
-                              !isValid ||
-                              !dirty ||
-                              hasUploadingFiles
+                          <SimpleTooltip
+                            tooltip={
+                              isSubmitting
+                                ? "Saving changes..."
+                                : !isValid
+                                  ? "Please fix the errors in the form before saving."
+                                  : !dirty
+                                    ? "No changes have been made."
+                                    : hasUploadingFiles
+                                      ? "Please wait for files to finish uploading."
+                                      : undefined
                             }
+                            side="bottom"
                           >
-                            <OpalButton type="submit">
-                              {existingAgent ? "Save" : "Create"}
-                            </OpalButton>
-                          </Disabled>
+                            <Disabled
+                              disabled={
+                                isSubmitting ||
+                                !isValid ||
+                                !dirty ||
+                                hasUploadingFiles
+                              }
+                            >
+                              <OpalButton type="submit">
+                                {existingAgent ? "Save" : "Create"}
+                              </OpalButton>
+                            </Disabled>
+                          </SimpleTooltip>
                         </div>
                       }
                       backButton

web/src/refresh-pages/AppPage.tsx

@@ -465,7 +465,6 @@ export default function AppPage({ firstMessage }: ChatPageProps) {
 
   const onChat = useCallback(
     (message: string) => {
-      resetInputBar();
       onSubmit({
         message,
         currentMessageFiles,
@@ -476,7 +475,6 @@ export default function AppPage({ firstMessage }: ChatPageProps) {
       }
     },
     [
-      resetInputBar,
       onSubmit,
       currentMessageFiles,
       deepResearchEnabledForCurrentWorkflow,
@@ -510,7 +508,6 @@ export default function AppPage({ firstMessage }: ChatPageProps) {
       // If we're in an existing chat session, always use chat mode
       // (appMode only applies to new sessions)
       if (currentChatSessionId) {
-        resetInputBar();
         onSubmit({
           message,
           currentMessageFiles,
@@ -523,7 +520,7 @@ export default function AppPage({ firstMessage }: ChatPageProps) {
       }
 
       // For new sessions, let the query controller handle routing.
-      // resetInputBar is called inside onChat for chat-routed queries.
+      // resetInputBar is called inside useChatController.onSubmit for chat-routed queries.
       // For search-routed queries, the input bar is intentionally kept
       // so the user can see and refine their search query.
       await submitQuery(message, onChat);
@@ -532,7 +529,6 @@ export default function AppPage({ firstMessage }: ChatPageProps) {
       currentChatSessionId,
       submitQuery,
       onChat,
-      resetInputBar,
       onSubmit,
       currentMessageFiles,
       deepResearchEnabledForCurrentWorkflow,

web/src/refresh-pages/admin/ChatPreferencesPage.tsx

@@ -114,6 +114,10 @@ function MCPServerCard({
   const allToolIds = tools.map((t) => t.id);
   const serverEnabled =
     tools.length > 0 && tools.some((t) => isToolEnabled(t.id));
+  const needsAuth = !server.is_authenticated;
+  const authTooltip = needsAuth
+    ? "Authenticate this MCP server before enabling its tools."
+    : undefined;
 
   return (
     <ExpandableCard.Root isFolded={isFolded} onFoldedChange={setIsFolded}>
@@ -122,10 +126,13 @@ function MCPServerCard({
         description={server.description}
         icon={getActionIcon(server.server_url, server.name)}
         rightChildren={
-          <Switch
-            checked={serverEnabled}
-            onCheckedChange={(checked) => onToggleTools(allToolIds, checked)}
-          />
+          <SimpleTooltip tooltip={authTooltip} side="top">
+            <Switch
+              checked={serverEnabled}
+              onCheckedChange={(checked) => onToggleTools(allToolIds, checked)}
+              disabled={needsAuth}
+            />
+          </SimpleTooltip>
         }
       >
         {tools.length > 0 && (
@@ -158,12 +165,15 @@ function MCPServerCard({
                 description={tool.description}
                 icon={tool.icon}
                 rightChildren={
-                  <Switch
-                    checked={isToolEnabled(tool.id)}
-                    onCheckedChange={(checked) =>
-                      onToggleTool(tool.id, checked)
-                    }
-                  />
+                  <SimpleTooltip tooltip={authTooltip} side="top">
+                    <Switch
+                      checked={isToolEnabled(tool.id)}
+                      onCheckedChange={(checked) =>
+                        onToggleTool(tool.id, checked)
+                      }
+                      disabled={needsAuth}
+                    />
+                  </SimpleTooltip>
                 }
               />
             ))}

web/src/sections/input/AppInputBar.tsx

@@ -21,6 +21,7 @@ import { ChatState } from "@/app/app/interfaces";
 import { useForcedTools } from "@/lib/hooks/useForcedTools";
 import { useAppMode } from "@/providers/AppModeProvider";
 import useAppFocus from "@/hooks/useAppFocus";
+import { getPastedFilesIfNoText } from "@/lib/clipboard";
 import { cn, isImageFile } from "@/lib/utils";
 import { Disabled } from "@opal/core";
 import { useUser } from "@/providers/UserProvider";
@@ -233,20 +234,10 @@ const AppInputBar = React.memo(
     }, [showFiles, currentMessageFiles]);
 
     function handlePaste(event: React.ClipboardEvent) {
-      const items = event.clipboardData?.items;
-      if (items) {
-        const pastedFiles = [];
-        for (let i = 0; i < items.length; i++) {
-          const item = items[i];
-          if (item && item.kind === "file") {
-            const file = item.getAsFile();
-            if (file) pastedFiles.push(file);
-          }
-        }
-        if (pastedFiles.length > 0) {
-          event.preventDefault();
-          handleFileUpload(pastedFiles);
-        }
+      const pastedFiles = getPastedFilesIfNoText(event.clipboardData);
+      if (pastedFiles.length > 0) {
+        event.preventDefault();
+        handleFileUpload(pastedFiles);
       }
     }
 

web/src/sections/modals/PreviewModal/PreviewModal.tsx

@@ -7,21 +7,16 @@ import Text from "@/refresh-components/texts/Text";
 import SimpleLoader from "@/refresh-components/loaders/SimpleLoader";
 import { cn } from "@/lib/utils";
 import { Section } from "@/layouts/general-layouts";
-import { getCodeLanguage, getDataLanguage } from "@/lib/languages";
+import mime from "mime";
+import {
+  getCodeLanguage,
+  getDataLanguage,
+  getLanguageByMime,
+} from "@/lib/languages";
 import { fetchChatFile } from "@/lib/chat/svc";
 import { PreviewContext } from "@/sections/modals/PreviewModal/interfaces";
 import { resolveVariant } from "@/sections/modals/PreviewModal/variants";
 
-function resolveMimeType(mimeType: string, fileName: string): string {
-  if (mimeType !== "application/octet-stream") return mimeType;
-  const lower = fileName.toLowerCase();
-  if (lower.endsWith(".md") || lower.endsWith(".markdown"))
-    return "text/markdown";
-  if (lower.endsWith(".txt")) return "text/plain";
-  if (lower.endsWith(".csv")) return "text/csv";
-  return mimeType;
-}
-
 interface PreviewModalProps {
   presentingDocument: MinimalOnyxDocument;
   onClose: () => void;
@@ -47,9 +42,10 @@ export default function PreviewModal({
   const language = useMemo(
     () =>
       getCodeLanguage(presentingDocument.semantic_identifier || "") ||
+      getLanguageByMime(mimeType) ||
       getDataLanguage(presentingDocument.semantic_identifier || "") ||
       "plaintext",
-    [presentingDocument.semantic_identifier]
+    [mimeType, presentingDocument.semantic_identifier]
   );
 
   const lineCount = useMemo(() => {
@@ -91,7 +87,10 @@ export default function PreviewModal({
 
       const rawContentType =
         response.headers.get("Content-Type") || "application/octet-stream";
-      const resolvedMime = resolveMimeType(rawContentType, originalFileName);
+      const resolvedMime =
+        rawContentType === "application/octet-stream"
+          ? mime.getType(originalFileName) ?? rawContentType
+          : rawContentType;
       setMimeType(resolvedMime);
 
       const resolved = resolveVariant(
@@ -171,24 +170,24 @@ export default function PreviewModal({
           onClose={onClose}
         />
 
-        {/* Body + floating footer wrapper */}
-        <Modal.Body padding={0} gap={0}>
-          <Section padding={0} gap={0}>
-            {isLoading ? (
-              <Section>
-                <SimpleLoader className="h-8 w-8" />
-              </Section>
-            ) : loadError ? (
-              <Section padding={1}>
-                <Text text03 mainUiBody>
-                  {loadError}
-                </Text>
-              </Section>
-            ) : (
-              variant.renderContent(ctx)
-            )}
-          </Section>
-        </Modal.Body>
+        {/* Body — uses flex-1/min-h-0/overflow-hidden (not Modal.Body)
+            so that child ScrollIndicatorDivs become the actual scroll
+            container instead of the body stealing it via overflow-y-auto. */}
+        <div className="flex flex-col flex-1 min-h-0 overflow-hidden w-full bg-background-tint-01">
+          {isLoading ? (
+            <Section>
+              <SimpleLoader className="h-8 w-8" />
+            </Section>
+          ) : loadError ? (
+            <Section padding={1}>
+              <Text text03 mainUiBody>
+                {loadError}
+              </Text>
+            </Section>
+          ) : (
+            variant.renderContent(ctx)
+          )}
+        </div>
 
         {/* Floating footer */}
         {!isLoading && !loadError && (
@@ -199,8 +198,9 @@ export default function PreviewModal({
               "p-4 pointer-events-none w-full"
             )}
             style={{
-              background:
-                "linear-gradient(to top, var(--background-tint-01) 40%, transparent)",
+              background: `linear-gradient(to top, var(--background-${
+                variant.codeBackground ? "code-01" : "tint-01"
+              }) 40%, transparent)`,
             }}
           >
             {/* Left slot */}

web/src/sections/modals/PreviewModal/interfaces.ts

@@ -19,6 +19,8 @@ export interface PreviewVariant
   matches: (semanticIdentifier: string | null, mimeType: string) => boolean;
   /** Whether the fetcher should read the blob as text. */
   needsTextContent: boolean;
+  /** Whether the variant renders on a code-style background (bg-background-code-01). */
+  codeBackground: boolean;
   /** String shown below the title in the modal header. */
   headerDescription: (ctx: PreviewContext) => string;
   /** Body content. */

web/src/sections/modals/PreviewModal/variants/CodePreview.tsx

@@ -0,0 +1,37 @@
+"use client";
+
+import MinimalMarkdown from "@/components/chat/MinimalMarkdown";
+import ScrollIndicatorDiv from "@/refresh-components/ScrollIndicatorDiv";
+import { cn } from "@/lib/utils";
+import "@/app/app/message/custom-code-styles.css";
+
+interface CodePreviewProps {
+  content: string;
+  language?: string | null;
+  normalize?: boolean;
+}
+
+export function CodePreview({
+  content,
+  language,
+  normalize,
+}: CodePreviewProps) {
+  // Wrap raw content in a fenced code block for syntax highlighting. Uses ~~~
+  // instead of ``` to avoid conflicts with backticks in the content. Any literal
+  // ~~~ sequences in the content are escaped so they don't accidentally close the fence.
+  const markdownContent = normalize
+    ? `~~~${language || ""}\n${content.replace(/~~~/g, "\\~\\~\\~")}\n~~~`
+    : content;
+
+  return (
+    <ScrollIndicatorDiv
+      className={cn("p-4", normalize && "bg-background-code-01")}
+      backgroundColor={normalize ? "var(--background-code-01)" : undefined}
+      variant="shadow"
+      bottomSpacing="2rem"
+      disableBottomIndicator
+    >
+      <MinimalMarkdown content={markdownContent} showHeader={false} />
+    </ScrollIndicatorDiv>
+  );
+}

web/src/sections/modals/PreviewModal/variants/codeVariant.tsx

@@ -1,10 +1,8 @@
-import MinimalMarkdown from "@/components/chat/MinimalMarkdown";
 import Text from "@/refresh-components/texts/Text";
 import { Section } from "@/layouts/general-layouts";
 import { getCodeLanguage } from "@/lib/languages";
-import { CodeBlock } from "@/app/app/message/CodeBlock";
-import { extractCodeText } from "@/app/app/message/codeUtils";
 import { PreviewVariant } from "@/sections/modals/PreviewModal/interfaces";
+import { CodePreview } from "@/sections/modals/PreviewModal/variants/CodePreview";
 import {
   CopyButton,
   DownloadButton,
@@ -15,6 +13,7 @@ export const codeVariant: PreviewVariant = {
   width: "md",
   height: "lg",
   needsTextContent: true,
+  codeBackground: true,
 
   headerDescription: (ctx) =>
     ctx.fileContent
@@ -24,20 +23,7 @@ export const codeVariant: PreviewVariant = {
       : "",
 
   renderContent: (ctx) => (
-    <MinimalMarkdown
-      content={`\`\`\`${ctx.language}\n${ctx.fileContent}\n\n\`\`\``}
-      className="w-full break-words h-full"
-      components={{
-        code: ({ node, children }: any) => {
-          const codeText = extractCodeText(node, ctx.fileContent, children);
-          return (
-            <CodeBlock className="" codeText={codeText}>
-              {children}
-            </CodeBlock>
-          );
-        },
-      }}
-    />
+    <CodePreview normalize content={ctx.fileContent} language={ctx.language} />
   ),
 
   renderFooterLeft: (ctx) => (

web/src/sections/modals/PreviewModal/variants/csvVariant.tsx

@@ -34,6 +34,7 @@ export const csvVariant: PreviewVariant = {
   width: "lg",
   height: "full",
   needsTextContent: true,
+  codeBackground: false,
   headerDescription: (ctx) => {
     if (!ctx.fileContent) return "";
     const { rows } = parseCsv(ctx.fileContent);

web/src/sections/modals/PreviewModal/variants/dataVariant.tsx

@@ -1,10 +1,8 @@
-import MinimalMarkdown from "@/components/chat/MinimalMarkdown";
 import Text from "@/refresh-components/texts/Text";
 import { Section } from "@/layouts/general-layouts";
-import { getDataLanguage } from "@/lib/languages";
-import { CodeBlock } from "@/app/app/message/CodeBlock";
-import { extractCodeText } from "@/app/app/message/codeUtils";
+import { getDataLanguage, getLanguageByMime } from "@/lib/languages";
 import { PreviewVariant } from "@/sections/modals/PreviewModal/interfaces";
+import { CodePreview } from "@/sections/modals/PreviewModal/variants/CodePreview";
 import {
   CopyButton,
   DownloadButton,
@@ -22,10 +20,12 @@ function formatContent(language: string, content: string): string {
 }
 
 export const dataVariant: PreviewVariant = {
-  matches: (name) => !!getDataLanguage(name || ""),
+  matches: (name, mime) =>
+    !!getDataLanguage(name || "") || !!getLanguageByMime(mime),
   width: "md",
   height: "lg",
   needsTextContent: true,
+  codeBackground: true,
 
   headerDescription: (ctx) =>
     ctx.fileContent
@@ -37,20 +37,7 @@ export const dataVariant: PreviewVariant = {
   renderContent: (ctx) => {
     const formatted = formatContent(ctx.language, ctx.fileContent);
     return (
-      <MinimalMarkdown
-        content={`\`\`\`${ctx.language}\n${formatted}\n\n\`\`\``}
-        className="w-full break-words h-full"
-        components={{
-          code: ({ node, children }: any) => {
-            const codeText = extractCodeText(node, formatted, children);
-            return (
-              <CodeBlock className="" codeText={codeText}>
-                {children}
-              </CodeBlock>
-            );
-          },
-        }}
-      />
+      <CodePreview normalize content={formatted} language={ctx.language} />
     );
   },
 

web/src/sections/modals/PreviewModal/variants/docxVariant.tsx

@@ -130,6 +130,7 @@ export const docxVariant: PreviewVariant = {
   width: "lg",
   height: "full",
   needsTextContent: false,
+  codeBackground: false,
   headerDescription: () => {
     if (lastDocxResult) {
       const count = lastDocxResult.wordCount;

web/src/sections/modals/PreviewModal/variants/imageVariant.tsx

@@ -11,6 +11,7 @@ export const imageVariant: PreviewVariant = {
   width: "lg",
   height: "full",
   needsTextContent: false,
+  codeBackground: false,
   headerDescription: () => "",
 
   renderContent: (ctx) => (

web/src/sections/modals/PreviewModal/variants/index.ts

@@ -5,6 +5,7 @@ import { pdfVariant } from "@/sections/modals/PreviewModal/variants/pdfVariant";
 import { csvVariant } from "@/sections/modals/PreviewModal/variants/csvVariant";
 import { markdownVariant } from "@/sections/modals/PreviewModal/variants/markdownVariant";
 import { dataVariant } from "@/sections/modals/PreviewModal/variants/dataVariant";
+import { textVariant } from "@/sections/modals/PreviewModal/variants/textVariant";
 import { unsupportedVariant } from "@/sections/modals/PreviewModal/variants/unsupportedVariant";
 import { docxVariant } from "@/sections/modals/PreviewModal/variants/docxVariant";
 
@@ -14,9 +15,10 @@ const PREVIEW_VARIANTS: PreviewVariant[] = [
   imageVariant,
   pdfVariant,
   csvVariant,
-  dataVariant,
   markdownVariant,
   docxVariant,
+  textVariant,
+  dataVariant,
 ];
 
 export function resolveVariant(

web/src/sections/modals/PreviewModal/variants/markdownVariant.tsx

@@ -1,8 +1,7 @@
-import MinimalMarkdown from "@/components/chat/MinimalMarkdown";
-import ScrollIndicatorDiv from "@/refresh-components/ScrollIndicatorDiv";
 import { Section } from "@/layouts/general-layouts";
 import { isMarkdownFile } from "@/lib/languages";
 import { PreviewVariant } from "@/sections/modals/PreviewModal/interfaces";
+import { CodePreview } from "@/sections/modals/PreviewModal/variants/CodePreview";
 import {
   CopyButton,
   DownloadButton,
@@ -11,7 +10,6 @@ import {
 const MARKDOWN_MIMES = [
   "text/markdown",
   "text/x-markdown",
-  "text/plain",
   "text/x-rst",
   "text/x-org",
 ];
@@ -24,15 +22,11 @@ export const markdownVariant: PreviewVariant = {
   width: "lg",
   height: "full",
   needsTextContent: true,
+  codeBackground: false,
   headerDescription: () => "",
 
   renderContent: (ctx) => (
-    <ScrollIndicatorDiv className="flex-1 min-h-0 p-4" variant="shadow">
-      <MinimalMarkdown
-        content={ctx.fileContent}
-        className="w-full pb-4 text-lg break-words"
-      />
-    </ScrollIndicatorDiv>
+    <CodePreview content={ctx.fileContent} language={ctx.language} />
   ),
 
   renderFooterLeft: () => null,

web/src/sections/modals/PreviewModal/variants/pdfVariant.tsx

@@ -7,6 +7,7 @@ export const pdfVariant: PreviewVariant = {
   width: "lg",
   height: "full",
   needsTextContent: false,
+  codeBackground: false,
   headerDescription: () => "",
 
   renderContent: (ctx) => (

web/src/sections/modals/PreviewModal/variants/textVariant.tsx

@@ -0,0 +1,55 @@
+import Text from "@/refresh-components/texts/Text";
+import { Section } from "@/layouts/general-layouts";
+import { PreviewVariant } from "@/sections/modals/PreviewModal/interfaces";
+import { CodePreview } from "@/sections/modals/PreviewModal/variants/CodePreview";
+import {
+  CopyButton,
+  DownloadButton,
+} from "@/sections/modals/PreviewModal/variants/shared";
+
+const TEXT_MIMES = [
+  "text/plain",
+  "text/x-log",
+  "text/x-config",
+  "text/tab-separated-values",
+];
+
+const TEXT_EXTENSIONS = [".txt", ".log", ".conf", ".tsv"];
+
+export const textVariant: PreviewVariant = {
+  matches: (name, mime) => {
+    if (TEXT_MIMES.some((supportedMime) => mime.startsWith(supportedMime))) {
+      return true;
+    }
+
+    const lowerName = (name || "").toLowerCase();
+    return TEXT_EXTENSIONS.some((extension) => lowerName.endsWith(extension));
+  },
+  width: "md",
+  height: "lg",
+  needsTextContent: true,
+  codeBackground: true,
+  headerDescription: (ctx) =>
+    ctx.fileContent
+      ? `${ctx.lineCount} ${ctx.lineCount === 1 ? "line" : "lines"} · ${
+          ctx.fileSize
+        }`
+      : "",
+
+  renderContent: (ctx) => (
+    <CodePreview normalize content={ctx.fileContent} language={ctx.language} />
+  ),
+
+  renderFooterLeft: (ctx) => (
+    <Text text03 mainUiBody className="select-none">
+      {ctx.lineCount} {ctx.lineCount === 1 ? "line" : "lines"}
+    </Text>
+  ),
+
+  renderFooterRight: (ctx) => (
+    <Section flexDirection="row" width="fit">
+      <CopyButton getText={() => ctx.fileContent} />
+      <DownloadButton fileUrl={ctx.fileUrl} fileName={ctx.fileName} />
+    </Section>
+  ),
+};

web/src/sections/modals/PreviewModal/variants/unsupportedVariant.tsx

@@ -5,13 +5,14 @@ import { DownloadButton } from "@/sections/modals/PreviewModal/variants/shared";
 
 export const unsupportedVariant: PreviewVariant = {
   matches: () => true,
-  width: "lg",
+  width: "md",
   height: "full",
   needsTextContent: false,
+  codeBackground: false,
   headerDescription: () => "",
 
   renderContent: (ctx) => (
-    <div className="flex flex-col items-center justify-center flex-1 min-h-0 gap-4 p-6">
+    <div className="flex flex-col items-center justify-center flex-1 w-full min-h-0 gap-4 p-6">
       <Text as="p" text03 mainUiBody>
         This file format is not supported for preview.
       </Text>

web/src/sections/modals/ShareAgentModal.test.tsx

@@ -0,0 +1,83 @@
+import React, { useEffect } from "react";
+import { render, screen, waitFor } from "@tests/setup/test-utils";
+import ShareAgentModal, { ShareAgentModalProps } from "./ShareAgentModal";
+import { useCreateModal } from "@/refresh-components/contexts/ModalContext";
+
+jest.mock("@/hooks/useShareableUsers", () => ({
+  __esModule: true,
+  default: jest.fn(() => ({ data: [] })),
+}));
+
+jest.mock("@/hooks/useShareableGroups", () => ({
+  __esModule: true,
+  default: jest.fn(() => ({ data: [] })),
+}));
+
+jest.mock("@/hooks/useAgents", () => ({
+  useAgent: jest.fn(() => ({ agent: null })),
+}));
+
+jest.mock("@/lib/hooks", () => ({
+  useLabels: jest.fn(() => ({
+    labels: [],
+    createLabel: jest.fn(),
+  })),
+}));
+
+function ModalHarness(props: ShareAgentModalProps) {
+  const modal = useCreateModal();
+
+  useEffect(() => {
+    modal.toggle(true);
+  }, [modal]);
+
+  return (
+    <modal.Provider>
+      <ShareAgentModal {...props} />
+    </modal.Provider>
+  );
+}
+
+function renderShareAgentModal(overrides: Partial<ShareAgentModalProps> = {}) {
+  const props: ShareAgentModalProps = {
+    userIds: [],
+    groupIds: [],
+    isPublic: false,
+    isFeatured: false,
+    labelIds: [],
+    ...overrides,
+  };
+
+  return render(<ModalHarness {...props} />);
+}
+
+describe("ShareAgentModal", () => {
+  it("defaults to Users & Groups when the agent is private", async () => {
+    renderShareAgentModal({ isPublic: false });
+
+    await waitFor(() =>
+      expect(
+        screen.getByRole("tab", { name: "Users & Groups" })
+      ).toHaveAttribute("data-state", "active")
+    );
+
+    expect(
+      screen.getByRole("tab", { name: "Your Organization" })
+    ).toHaveAttribute("data-state", "inactive");
+  });
+
+  it("defaults to Your Organization when the agent is public", async () => {
+    renderShareAgentModal({ isPublic: true });
+
+    await waitFor(() =>
+      expect(
+        screen.getByRole("tab", { name: "Your Organization" })
+      ).toHaveAttribute("data-state", "active")
+    );
+
+    expect(screen.getByRole("tab", { name: "Users & Groups" })).toHaveAttribute(
+      "data-state",
+      "inactive"
+    );
+  });
+});

web/src/sections/modals/TextViewModal.tsx

@@ -1,333 +0,0 @@
-"use client";
-
-import { useState, useEffect, useCallback, useMemo } from "react";
-import {
-  Table,
-  TableBody,
-  TableCell,
-  TableHead,
-  TableHeader,
-  TableRow,
-} from "@/components/ui/table";
-import { MinimalOnyxDocument } from "@/lib/search/interfaces";
-import MinimalMarkdown from "@/components/chat/MinimalMarkdown";
-import { Button } from "@opal/components";
-import Modal, { BasicModalFooter } from "@/refresh-components/Modal";
-import Text from "@/refresh-components/texts/Text";
-import {
-  SvgDownloadCloud,
-  SvgFileText,
-  SvgZoomIn,
-  SvgZoomOut,
-} from "@opal/icons";
-import PreviewImage from "@/refresh-components/PreviewImage";
-import SimpleLoader from "@/refresh-components/loaders/SimpleLoader";
-import ScrollIndicatorDiv from "@/refresh-components/ScrollIndicatorDiv";
-import { cn } from "@/lib/utils";
-import { Section } from "@/layouts/general-layouts";
-
-export interface TextViewProps {
-  presentingDocument: MinimalOnyxDocument;
-  onClose: () => void;
-}
-
-export default function TextViewModal({
-  presentingDocument,
-  onClose,
-}: TextViewProps) {
-  const [zoom, setZoom] = useState(100);
-  const [fileContent, setFileContent] = useState("");
-  const [fileUrl, setFileUrl] = useState("");
-  const [fileName, setFileName] = useState("");
-  const [isLoading, setIsLoading] = useState(true);
-  const [loadError, setLoadError] = useState<string | null>(null);
-  const [fileType, setFileType] = useState("application/octet-stream");
-  const csvData = useMemo(() => {
-    if (!fileType.startsWith("text/csv")) {
-      return null;
-    }
-
-    const lines = fileContent.split(/\r?\n/).filter((l) => l.length > 0);
-    const headers = lines.length > 0 ? lines[0]?.split(",") ?? [] : [];
-    const rows = lines.slice(1).map((line) => line.split(","));
-
-    return { headers, rows } as { headers: string[]; rows: string[][] };
-  }, [fileContent, fileType]);
-
-  // Detect if a given MIME type is one of the recognized markdown formats
-  const isMarkdownFormat = (mimeType: string): boolean => {
-    const markdownFormats = [
-      "text/markdown",
-      "text/x-markdown",
-      "text/plain",
-      "text/csv",
-      "text/x-rst",
-      "text/x-org",
-      "txt",
-    ];
-    return markdownFormats.some((format) => mimeType.startsWith(format));
-  };
-
-  const isImageFormat = (mimeType: string) => {
-    const imageFormats = [
-      "image/png",
-      "image/jpeg",
-      "image/gif",
-      "image/svg+xml",
-    ];
-    return imageFormats.some((format) => mimeType.startsWith(format));
-  };
-  // Detect if a given MIME type can be rendered in an <iframe>
-  const isSupportedIframeFormat = (mimeType: string): boolean => {
-    const supportedFormats = [
-      "application/pdf",
-      "image/png",
-      "image/jpeg",
-      "image/gif",
-      "image/svg+xml",
-    ];
-    return supportedFormats.some((format) => mimeType.startsWith(format));
-  };
-
-  const fetchFile = useCallback(
-    async (signal?: AbortSignal) => {
-      setIsLoading(true);
-      setLoadError(null);
-      setFileContent("");
-      const fileIdLocal =
-        presentingDocument.document_id.split("__")[1] ||
-        presentingDocument.document_id;
-
-      try {
-        const response = await fetch(
-          `/api/chat/file/${encodeURIComponent(fileIdLocal)}`,
-          {
-            method: "GET",
-            signal,
-            cache: "force-cache",
-          }
-        );
-
-        if (!response.ok) {
-          setLoadError("Failed to load document.");
-          return;
-        }
-
-        const blob = await response.blob();
-        const url = window.URL.createObjectURL(blob);
-        setFileUrl((prev) => {
-          if (prev) {
-            window.URL.revokeObjectURL(prev);
-          }
-          return url;
-        });
-
-        const originalFileName =
-          presentingDocument.semantic_identifier || "document";
-        setFileName(originalFileName);
-
-        let contentType =
-          response.headers.get("Content-Type") || "application/octet-stream";
-
-        // If it's octet-stream but file name suggests a text-based extension, override accordingly
-        if (contentType === "application/octet-stream") {
-          const lowerName = originalFileName.toLowerCase();
-          if (lowerName.endsWith(".md") || lowerName.endsWith(".markdown")) {
-            contentType = "text/markdown";
-          } else if (lowerName.endsWith(".txt")) {
-            contentType = "text/plain";
-          } else if (lowerName.endsWith(".csv")) {
-            contentType = "text/csv";
-          }
-        }
-        setFileType(contentType);
-
-        // If the final content type looks like markdown, read its text
-        if (isMarkdownFormat(contentType)) {
-          const text = await blob.text();
-          setFileContent(text);
-        }
-      } catch (error) {
-        // Abort is expected on unmount / doc change
-        if (signal?.aborted) {
-          return;
-        }
-        setLoadError("Failed to load document.");
-      } finally {
-        // Prevent stale/aborted requests from clobbering the loading state.
-        // This is especially important in React StrictMode where effects can run twice.
-        if (!signal?.aborted) {
-          setIsLoading(false);
-        }
-      }
-    },
-    [presentingDocument]
-  );
-
-  useEffect(() => {
-    const controller = new AbortController();
-    fetchFile(controller.signal);
-    return () => {
-      controller.abort();
-    };
-  }, [fetchFile]);
-
-  useEffect(() => {
-    return () => {
-      if (fileUrl) {
-        window.URL.revokeObjectURL(fileUrl);
-      }
-    };
-  }, [fileUrl]);
-
-  const handleDownload = () => {
-    const link = document.createElement("a");
-    link.href = fileUrl;
-    link.download = fileName || presentingDocument.document_id;
-    document.body.appendChild(link);
-    link.click();
-    document.body.removeChild(link);
-  };
-
-  const handleZoomIn = () => setZoom((prev) => Math.min(prev + 25, 200));
-  const handleZoomOut = () => setZoom((prev) => Math.max(prev - 25, 100));
-
-  return (
-    <Modal
-      open
-      onOpenChange={(open) => {
-        if (!open) {
-          onClose();
-        }
-      }}
-    >
-      <Modal.Content
-        width="lg"
-        height="full"
-        preventAccidentalClose={false}
-        onOpenAutoFocus={(e) => e.preventDefault()}
-      >
-        <Modal.Header
-          icon={SvgFileText}
-          title={fileName || "Document"}
-          onClose={onClose}
-        >
-          <Section flexDirection="row" justifyContent="start" gap={0.25}>
-            <Button
-              prominence="tertiary"
-              onClick={handleZoomOut}
-              icon={SvgZoomOut}
-              tooltip="Zoom Out"
-            />
-            <Text mainUiBody>{zoom}%</Text>
-            <Button
-              prominence="tertiary"
-              onClick={handleZoomIn}
-              icon={SvgZoomIn}
-              tooltip="Zoom In"
-            />
-            <Button
-              prominence="tertiary"
-              onClick={handleDownload}
-              icon={SvgDownloadCloud}
-              tooltip="Download"
-            />
-          </Section>
-        </Modal.Header>
-
-        <Modal.Body>
-          <Section>
-            {isLoading ? (
-              <SimpleLoader className="h-8 w-8" />
-            ) : loadError ? (
-              <Text text03 mainUiBody>
-                {loadError}
-              </Text>
-            ) : (
-              <div
-                className="flex flex-col flex-1 min-h-0 min-w-0 w-full transform origin-center transition-transform duration-300 ease-in-out"
-                style={{ transform: `scale(${zoom / 100})` }}
-              >
-                {isImageFormat(fileType) ? (
-                  <PreviewImage
-                    src={fileUrl}
-                    alt={fileName}
-                    className="w-full flex-1 min-h-0"
-                  />
-                ) : isSupportedIframeFormat(fileType) ? (
-                  <iframe
-                    src={`${fileUrl}#toolbar=0`}
-                    className="w-full h-full flex-1 min-h-0 border-none"
-                    title="File Viewer"
-                  />
-                ) : isMarkdownFormat(fileType) ? (
-                  <ScrollIndicatorDiv
-                    className="flex-1 min-h-0 p-4"
-                    variant="shadow"
-                  >
-                    {csvData ? (
-                      <Table>
-                        <TableHeader className="sticky top-0 z-sticky">
-                          <TableRow className="bg-background-tint-02">
-                            {csvData.headers.map((h, i) => (
-                              <TableHead key={i}>
-                                <Text
-                                  as="p"
-                                  className="line-clamp-2 font-medium"
-                                  text03
-                                  mainUiBody
-                                >
-                                  {h}
-                                </Text>
-                              </TableHead>
-                            ))}
-                          </TableRow>
-                        </TableHeader>
-                        <TableBody>
-                          {csvData.rows.map((row, rIdx) => (
-                            <TableRow key={rIdx}>
-                              {csvData.headers.map((_, cIdx) => (
-                                <TableCell
-                                  key={cIdx}
-                                  className={cn(
-                                    cIdx === 0 &&
-                                      "sticky left-0 bg-background-tint-01",
-                                    "py-0 px-4 whitespace-normal break-words"
-                                  )}
-                                >
-                                  {row?.[cIdx] ?? ""}
-                                </TableCell>
-                              ))}
-                            </TableRow>
-                          ))}
-                        </TableBody>
-                      </Table>
-                    ) : (
-                      <MinimalMarkdown
-                        content={fileContent}
-                        className="w-full pb-4 h-full text-lg break-words"
-                      />
-                    )}
-                  </ScrollIndicatorDiv>
-                ) : (
-                  <div className="flex flex-col items-center justify-center flex-1 min-h-0 p-6 gap-4">
-                    <Text as="p" text03 mainUiBody>
-                      This file format is not supported for preview.
-                    </Text>
-                    <Button onClick={handleDownload}>Download File</Button>
-                  </div>
-                )}
-              </div>
-            )}
-          </Section>
-        </Modal.Body>
-
-        <Modal.Footer>
-          <BasicModalFooter
-            submit={<Button onClick={handleDownload}>Download File</Button>}
-          />
-        </Modal.Footer>
-      </Modal.Content>
-    </Modal>
-  );
-}

web/tailwind-themes/tailwind.config.js

@@ -260,6 +260,7 @@ module.exports = {
         "code-string": "var(--code-string)",
         "code-number": "var(--code-number)",
         "code-definition": "var(--code-definition)",
+        "background-code-01": "var(--background-code-01)",
 
         // Shimmer colors for loading animations
         "shimmer-base": "var(--shimmer-base)",

web/tests/e2e/admin/admin_pages.spec.ts

@@ -187,7 +187,10 @@ for (const theme of THEMES) {
           /\//g,
           "-"
         )}`;
-        await expectScreenshot(page, { name: screenshotName });
+        await expectScreenshot(page, {
+          name: screenshotName,
+          mask: ['[data-testid="admin-date-range-selector-button"]'],
+        });
       });
     }
   });

web/tests/e2e/mcp/mcp_oauth_flow.spec.ts

@@ -1860,6 +1860,9 @@ test.describe("MCP OAuth flows", () => {
       toolName: TOOL_NAMES.admin,
       logStep,
     });
+    const createdAgent = await adminApiClient.getAssistant(agentId);
+    expect(createdAgent.is_public).toBe(false);
+    logStep("Verified newly created agent is private by default");
     const adminToolId = await fetchMcpToolIdByName(
       page,
       serverId,
@@ -1899,6 +1902,13 @@ test.describe("MCP OAuth flows", () => {
     ).toBeVisible({ timeout: 15000 });
     logStep("Verified MCP server card is still visible on actions page");
 
+    await adminApiClient.updateAgentSharing(agentId, {
+      isPublic: true,
+      userIds: createdAgent.users.map((user) => user.id),
+      groupIds: createdAgent.groups,
+    });
+    logStep("Published agent explicitly for end-user MCP flow");
+
     adminArtifacts = {
       serverId,
       serverName,

web/tests/e2e/utils/onyxApiClient.ts

@@ -681,6 +681,9 @@ export class OnyxApiClient {
 
   async getAssistant(agentId: number): Promise<{
     id: number;
+    is_public: boolean;
+    users: Array<{ id: string }>;
+    groups: number[];
     tools: Array<{ id: number; mcp_server_id?: number | null }>;
   }> {
     const response = await this.get(`/persona/${agentId}`);
@@ -690,6 +693,37 @@ export class OnyxApiClient {
     );
   }
 
+  async updateAgentSharing(
+    agentId: number,
+    options: {
+      userIds?: string[];
+      groupIds?: number[];
+      isPublic?: boolean;
+      labelIds?: number[];
+    }
+  ): Promise<void> {
+    const response = await this.request.patch(
+      `${this.baseUrl}/persona/${agentId}/share`,
+      {
+        data: {
+          user_ids: options.userIds,
+          group_ids: options.groupIds,
+          is_public: options.isPublic,
+          label_ids: options.labelIds,
+        },
+      }
+    );
+    await this.handleResponse(
+      response,
+      `Failed to update sharing for assistant ${agentId}`
+    );
+    this.log(
+      `Updated assistant sharing: ${agentId} (is_public=${String(
+        options.isPublic
+      )})`
+    );
+  }
+
   async listMcpServers(): Promise<any[]> {
     const response = await this.get(`/admin/mcp/servers`);
     const data = await this.handleResponse<{ mcp_servers: any[] }>(