k

.github/workflows/docker-build-backend-container-on-merge-group.yml

@@ -1,8 +1,6 @@
 name: Build Backend Image on Merge Group
 
 on:
-  pull_request:
-    branches: [ "main" ]
   merge_group:
     types: [checks_requested]
 

.github/workflows/docker-build-web-container-on-merge-group.yml

@@ -1,8 +1,6 @@
 name: Build Web Image on Merge Group
 
 on:
-  pull_request:
-    branches: [ "main" ]
   merge_group:
     types: [checks_requested]
 

backend/alembic/versions/05c07bf07c00_add_search_doc_relevance_details.py

@@ -1,7 +1,7 @@
 """add search doc relevance details
 
 Revision ID: 05c07bf07c00
-Revises: 3a7802814195
+Revises: b896bbd0d5a7
 Create Date: 2024-07-10 17:48:15.886653
 
 """

backend/alembic/versions/b896bbd0d5a7_backfill_is_internet_data_to_false.py

@@ -11,8 +11,8 @@ from alembic import op
 # revision identifiers, used by Alembic.
 revision = "b896bbd0d5a7"
 down_revision = "44f856ae2a4a"
-branch_labels = None
-depends_on = None
+branch_labels: None = None
+depends_on: None = None
 
 
 def upgrade() -> None:

backend/danswer/configs/app_configs.py

@@ -223,6 +223,11 @@ MAX_PRUNING_DOCUMENT_RETRIEVAL_PER_MINUTE = int(
     os.environ.get("MAX_PRUNING_DOCUMENT_RETRIEVAL_PER_MINUTE", 0)
 )
 
+# comma delimited list of zendesk article labels to skip indexing for
+ZENDESK_CONNECTOR_SKIP_ARTICLE_LABELS = os.environ.get(
+    "ZENDESK_CONNECTOR_SKIP_ARTICLE_LABELS", ""
+).split(",")
+
 
 #####
 # Indexing Configs

backend/danswer/connectors/confluence/rate_limit_handler.py

@@ -38,7 +38,7 @@ def make_confluence_call_handle_rate_limit(confluence_call: F) -> F:
                     retry_after = None
                     try:
                         retry_after = int(e.response.headers.get("Retry-After"))
-                    except ValueError:
+                    except (ValueError, TypeError):
                         pass
 
                     if retry_after:

backend/danswer/connectors/file/connector.py

@@ -88,7 +88,7 @@ def _process_file(
     # add a prefix to avoid conflicts with other connectors
     doc_id = f"FILE_CONNECTOR__{file_name}"
     if metadata:
-        doc_id = metadata.get("id") or doc_id
+        doc_id = metadata.get("document_id") or doc_id
 
     # If this is set, we will show this in the UI as the "name" of the file
     file_display_name = all_metadata.get("file_display_name") or os.path.basename(
@@ -111,6 +111,7 @@ def _process_file(
         for k, v in all_metadata.items()
         if k
         not in [
+            "document_id",
             "time_updated",
             "doc_updated_at",
             "link",

backend/danswer/connectors/zendesk/connector.py

@@ -4,6 +4,7 @@ from zenpy import Zenpy  # type: ignore
 from zenpy.lib.api_objects.help_centre_objects import Article  # type: ignore
 
 from danswer.configs.app_configs import INDEX_BATCH_SIZE
+from danswer.configs.app_configs import ZENDESK_CONNECTOR_SKIP_ARTICLE_LABELS
 from danswer.configs.constants import DocumentSource
 from danswer.connectors.cross_connector_utils.miscellaneous_utils import (
     time_str_to_utc,
@@ -81,7 +82,14 @@ class ZendeskConnector(LoadConnector, PollConnector):
         )
         doc_batch = []
         for article in articles:
-            if article.body is None or article.draft:
+            if (
+                article.body is None
+                or article.draft
+                or any(
+                    label in ZENDESK_CONNECTOR_SKIP_ARTICLE_LABELS
+                    for label in article.label_names
+                )
+            ):
                 continue
 
             doc_batch.append(_article_to_document(article))

backend/danswer/db/chat.py

@@ -146,6 +146,12 @@ def delete_search_doc_message_relationship(
     db_session.commit()
 
 
+def delete_tool_call_for_message_id(message_id: int, db_session: Session) -> None:
+    stmt = delete(ToolCall).where(ToolCall.message_id == message_id)
+    db_session.execute(stmt)
+    db_session.commit()
+
+
 def delete_orphaned_search_docs(db_session: Session) -> None:
     orphaned_docs = (
         db_session.query(SearchDoc)
@@ -169,6 +175,7 @@ def delete_messages_and_files_from_chat_session(
     ).fetchall()
 
     for id, files in messages_with_files:
+        delete_tool_call_for_message_id(message_id=id, db_session=db_session)
         delete_search_doc_message_relationship(message_id=id, db_session=db_session)
         for file_info in files or {}:
             lobj_name = file_info.get("id")

backend/danswer/document_index/vespa/app_config/schemas/danswer_chunk.sd

@@ -153,43 +153,41 @@ schema DANSWER_CHUNK_NAME {
             query(query_embedding) tensor<float>(x[VARIABLE_DIM])
         }
 
-        # This must be separate function for normalize_linear to work
-        function vector_score() {
+        function title_vector_score() {
             expression {
                 # If no title, the full vector score comes from the content embedding
-                (query(title_content_ratio) * if(attribute(skip_title), closeness(field, embeddings), closeness(field, title_embedding))) +
-                ((1 - query(title_content_ratio)) * closeness(field, embeddings))
-            }
-        }
-
-        # This must be separate function for normalize_linear to work
-        function keyword_score() {
-            expression {
-                (query(title_content_ratio) * bm25(title)) +
-                ((1 - query(title_content_ratio)) * bm25(content))
+                #query(title_content_ratio) * if(attribute(skip_title), closeness(field, embeddings), closeness(field, title_embedding))
+                if(attribute(skip_title), closeness(field, embeddings), closeness(field, title_embedding))
             }
         }
 
         first-phase {
-            expression: vector_score
+            expression: closeness(field, embeddings)
         }
 
         # Weighted average between Vector Search and BM-25
-        # Each is a weighted average between the Title and Content fields
-        # Finally each doc is boosted by it's user feedback based boost and recency
-        # If any embedding or index field is missing, it just receives a score of 0
-        # Assumptions:
-        # - For a given query + corpus, the BM-25 scores will be relatively similar in distribution
-        #   therefore not normalizing before combining.
-        # - For documents without title, it gets a score of 0 for that and this is ok as documents
-        #   without any title match should be penalized.
         global-phase {
             expression {
                 (
                     # Weighted Vector Similarity Score
-                    (query(alpha) * normalize_linear(vector_score)) +
+                    (
+                        query(alpha) * (
+                            (query(title_content_ratio) * normalize_linear(title_vector_score))
+                            +
+                            ((1 - query(title_content_ratio)) * normalize_linear(closeness(field, embeddings)))
+                        )
+                    )
+
+                    +
+
                     # Weighted Keyword Similarity Score
-                    ((1 - query(alpha)) * normalize_linear(keyword_score))
+                    (
+                        (1 - query(alpha)) * (
+                            (query(title_content_ratio) * normalize_linear(bm25(title)))
+                            +
+                            ((1 - query(title_content_ratio)) * normalize_linear(bm25(content)))
+                        )
+                    )
                 )
                 # Boost based on user feedback
                 * document_boost
@@ -204,8 +202,6 @@ schema DANSWER_CHUNK_NAME {
             bm25(content)
             closeness(field, title_embedding)
             closeness(field, embeddings)
-            keyword_score
-            vector_score
             document_boost
             recency_bias
             closest(embeddings)

backend/danswer/llm/chat_llm.py

@@ -232,32 +232,6 @@ class DefaultMultiLLM(LLM):
 
         self._model_kwargs = model_kwargs
 
-    @staticmethod
-    def _log_prompt(prompt: LanguageModelInput) -> None:
-        if isinstance(prompt, list):
-            for ind, msg in enumerate(prompt):
-                if isinstance(msg, AIMessageChunk):
-                    if msg.content:
-                        log_msg = msg.content
-                    elif msg.tool_call_chunks:
-                        log_msg = "Tool Calls: " + str(
-                            [
-                                {
-                                    key: value
-                                    for key, value in tool_call.items()
-                                    if key != "index"
-                                }
-                                for tool_call in msg.tool_call_chunks
-                            ]
-                        )
-                    else:
-                        log_msg = ""
-                    logger.debug(f"Message {ind}:\n{log_msg}")
-                else:
-                    logger.debug(f"Message {ind}:\n{msg.content}")
-        if isinstance(prompt, str):
-            logger.debug(f"Prompt:\n{prompt}")
-
     def log_model_configs(self) -> None:
         logger.info(f"Config: {self.config}")
 
@@ -311,7 +285,7 @@ class DefaultMultiLLM(LLM):
             api_version=self._api_version,
         )
 
-    def invoke(
+    def _invoke_implementation(
         self,
         prompt: LanguageModelInput,
         tools: list[dict] | None = None,
@@ -319,7 +293,6 @@ class DefaultMultiLLM(LLM):
     ) -> BaseMessage:
         if LOG_DANSWER_MODEL_INTERACTIONS:
             self.log_model_configs()
-            self._log_prompt(prompt)
 
         response = cast(
             litellm.ModelResponse, self._completion(prompt, tools, tool_choice, False)
@@ -328,7 +301,7 @@ class DefaultMultiLLM(LLM):
             response.choices[0].message
         )
 
-    def stream(
+    def _stream_implementation(
         self,
         prompt: LanguageModelInput,
         tools: list[dict] | None = None,
@@ -336,7 +309,6 @@ class DefaultMultiLLM(LLM):
     ) -> Iterator[BaseMessage]:
         if LOG_DANSWER_MODEL_INTERACTIONS:
             self.log_model_configs()
-            self._log_prompt(prompt)
 
         if DISABLE_LITELLM_STREAMING:
             yield self.invoke(prompt)

backend/danswer/llm/custom_llm.py

@@ -76,7 +76,7 @@ class CustomModelServer(LLM):
     def log_model_configs(self) -> None:
         logger.debug(f"Custom model at: {self._endpoint}")
 
-    def invoke(
+    def _invoke_implementation(
         self,
         prompt: LanguageModelInput,
         tools: list[dict] | None = None,
@@ -84,7 +84,7 @@ class CustomModelServer(LLM):
     ) -> BaseMessage:
         return self._execute(prompt)
 
-    def stream(
+    def _stream_implementation(
         self,
         prompt: LanguageModelInput,
         tools: list[dict] | None = None,

backend/danswer/llm/interfaces.py

@@ -3,9 +3,12 @@ from collections.abc import Iterator
 from typing import Literal
 
 from langchain.schema.language_model import LanguageModelInput
+from langchain_core.messages import AIMessageChunk
 from langchain_core.messages import BaseMessage
 from pydantic import BaseModel
 
+from danswer.configs.app_configs import DISABLE_GENERATIVE_AI
+from danswer.configs.app_configs import LOG_DANSWER_MODEL_INTERACTIONS
 from danswer.utils.logger import setup_logger
 
 
@@ -23,6 +26,32 @@ class LLMConfig(BaseModel):
     api_version: str | None
 
 
+def log_prompt(prompt: LanguageModelInput) -> None:
+    if isinstance(prompt, list):
+        for ind, msg in enumerate(prompt):
+            if isinstance(msg, AIMessageChunk):
+                if msg.content:
+                    log_msg = msg.content
+                elif msg.tool_call_chunks:
+                    log_msg = "Tool Calls: " + str(
+                        [
+                            {
+                                key: value
+                                for key, value in tool_call.items()
+                                if key != "index"
+                            }
+                            for tool_call in msg.tool_call_chunks
+                        ]
+                    )
+                else:
+                    log_msg = ""
+                logger.debug(f"Message {ind}:\n{log_msg}")
+            else:
+                logger.debug(f"Message {ind}:\n{msg.content}")
+    if isinstance(prompt, str):
+        logger.debug(f"Prompt:\n{prompt}")
+
+
 class LLM(abc.ABC):
     """Mimics the LangChain LLM / BaseChatModel interfaces to make it easy
     to use these implementations to connect to a variety of LLM providers."""
@@ -45,20 +74,48 @@ class LLM(abc.ABC):
     def log_model_configs(self) -> None:
         raise NotImplementedError
 
-    @abc.abstractmethod
+    def _precall(self, prompt: LanguageModelInput) -> None:
+        if DISABLE_GENERATIVE_AI:
+            raise Exception("Generative AI is disabled")
+        if LOG_DANSWER_MODEL_INTERACTIONS:
+            log_prompt(prompt)
+
     def invoke(
         self,
         prompt: LanguageModelInput,
         tools: list[dict] | None = None,
         tool_choice: ToolChoiceOptions | None = None,
+    ) -> BaseMessage:
+        self._precall(prompt)
+        # TODO add a postcall to log model outputs independent of concrete class
+        # implementation
+        return self._invoke_implementation(prompt, tools, tool_choice)
+
+    @abc.abstractmethod
+    def _invoke_implementation(
+        self,
+        prompt: LanguageModelInput,
+        tools: list[dict] | None = None,
+        tool_choice: ToolChoiceOptions | None = None,
     ) -> BaseMessage:
         raise NotImplementedError
 
-    @abc.abstractmethod
     def stream(
         self,
         prompt: LanguageModelInput,
         tools: list[dict] | None = None,
         tool_choice: ToolChoiceOptions | None = None,
+    ) -> Iterator[BaseMessage]:
+        self._precall(prompt)
+        # TODO add a postcall to log model outputs independent of concrete class
+        # implementation
+        return self._stream_implementation(prompt, tools, tool_choice)
+
+    @abc.abstractmethod
+    def _stream_implementation(
+        self,
+        prompt: LanguageModelInput,
+        tools: list[dict] | None = None,
+        tool_choice: ToolChoiceOptions | None = None,
     ) -> Iterator[BaseMessage]:
         raise NotImplementedError

backend/danswer/llm/llm_provider_options.py

@@ -26,6 +26,7 @@ OPENAI_PROVIDER_NAME = "openai"
 OPEN_AI_MODEL_NAMES = [
     "gpt-4",
     "gpt-4o",
+    "gpt-4o-mini",
     "gpt-4-turbo",
     "gpt-4-turbo-preview",
     "gpt-4-1106-preview",

backend/danswer/one_shot_answer/answer_question.py

@@ -189,7 +189,7 @@ def stream_answer_objects(
         chunks_below=query_req.chunks_below,
         full_doc=query_req.full_doc,
         bypass_acl=bypass_acl,
-        evaluate_response=query_req.evaluate_response,
+        llm_doc_eval=query_req.llm_doc_eval,
     )
 
     answer_config = AnswerStyleConfig(

backend/danswer/one_shot_answer/models.py

@@ -37,7 +37,7 @@ class DirectQARequest(ChunkContext):
     # This is to toggle agentic evaluation:
     # 1. Evaluates whether each response is relevant or not
     # 2. Provides a summary of the document's relevance in the resulsts
-    evaluate_response: bool = False
+    llm_doc_eval: bool = False
     # If True, skips generative an AI response to the search query
     skip_gen_ai_answer_generation: bool = False
 

backend/danswer/tools/search/search_tool.py

@@ -39,7 +39,7 @@ SEARCH_RESPONSE_SUMMARY_ID = "search_response_summary"
 SEARCH_DOC_CONTENT_ID = "search_doc_content"
 SECTION_RELEVANCE_LIST_ID = "section_relevance_list"
 FINAL_CONTEXT_DOCUMENTS = "final_context_documents"
-SEARCH_EVALUATION_ID = "evaluate_response"
+SEARCH_EVALUATION_ID = "llm_doc_eval"
 
 
 class SearchResponseSummary(BaseModel):
@@ -85,7 +85,7 @@ class SearchTool(Tool):
         chunks_below: int = 0,
         full_doc: bool = False,
         bypass_acl: bool = False,
-        evaluate_response: bool = False,
+        llm_doc_eval: bool = False,
     ) -> None:
         self.user = user
         self.persona = persona
@@ -102,7 +102,7 @@ class SearchTool(Tool):
         self.full_doc = full_doc
         self.bypass_acl = bypass_acl
         self.db_session = db_session
-        self.evaluate_response = evaluate_response
+        self.llm_doc_eval = llm_doc_eval
 
     @property
     def name(self) -> str:
@@ -295,7 +295,7 @@ class SearchTool(Tool):
 
         yield ToolResponse(id=FINAL_CONTEXT_DOCUMENTS, response=llm_docs)
 
-        if self.evaluate_response and not DISABLE_AGENTIC_SEARCH:
+        if self.llm_doc_eval and not DISABLE_AGENTIC_SEARCH:
             yield ToolResponse(
                 id=SEARCH_EVALUATION_ID, response=search_pipeline.relevance_summaries
             )

backend/danswer/tools/utils.py

@@ -6,7 +6,13 @@ from danswer.llm.utils import get_default_llm_tokenizer
 from danswer.tools.tool import Tool
 
 
-OPEN_AI_TOOL_CALLING_MODELS = {"gpt-3.5-turbo", "gpt-4-turbo", "gpt-4"}
+OPEN_AI_TOOL_CALLING_MODELS = {
+    "gpt-3.5-turbo",
+    "gpt-4-turbo",
+    "gpt-4",
+    "gpt-4o",
+    "gpt-4o-mini",
+}
 
 
 def explicit_tool_calling_supported(model_provider: str, model_name: str) -> bool:

backend/scripts/api_inference_sample.py

@@ -33,6 +33,7 @@ def process_question(danswer_url: str, question: str, api_key: str | None) -> No
         "message": question,
         "chat_session_id": chat_session_id,
         "parent_message_id": None,
+        "file_descriptors": [],
         # Default Question Answer prompt
         "prompt_id": 0,
         # Not specifying any specific docs to chat to, we want to run a search

backend/scripts/test-openapi-key.py

@@ -2,6 +2,8 @@ from openai import OpenAI
 
 
 VALID_MODEL_LIST = [
+    "gpt-4o-mini",
+    "gpt-4o",
     "gpt-4-1106-preview",
     "gpt-4-vision-preview",
     "gpt-4",

backend/tests/regression/answer_quality/README.md

@@ -70,7 +70,7 @@ Edit `search_test_config.yaml` to set:
 - model_server_port
     - This is the port of the remote model server
     - Only need to set this if use_cloud_gpu is true
-- existing_test_suffix
+- existing_test_suffix (THIS IS NOT A SUFFIX ANYMORE, TODO UPDATE THE DOCS HERE)
     - Use this if you would like to relaunch a previous test instance
     - Input the suffix of the test you'd like to re-launch 
     - (E.g. to use the data from folder "test-1234-5678" put "-1234-5678")

backend/tests/regression/answer_quality/cli_utils.py

@@ -97,7 +97,7 @@ def get_docker_container_env_vars(suffix: str) -> dict:
 
 def manage_data_directories(suffix: str, base_path: str, use_cloud_gpu: bool) -> None:
     # Use the user's home directory as the base path
-    target_path = os.path.join(os.path.expanduser(base_path), f"test{suffix}")
+    target_path = os.path.join(os.path.expanduser(base_path), suffix)
     directories = {
         "DANSWER_POSTGRES_DATA_DIR": os.path.join(target_path, "postgres/"),
         "DANSWER_VESPA_DATA_DIR": os.path.join(target_path, "vespa/"),
@@ -144,26 +144,30 @@ def _is_port_in_use(port: int) -> bool:
 
 
 def start_docker_compose(
-    run_suffix: str, launch_web_ui: bool, use_cloud_gpu: bool
+    run_suffix: str, launch_web_ui: bool, use_cloud_gpu: bool, only_state: bool = False
 ) -> None:
     print("Starting Docker Compose...")
     os.chdir(os.path.dirname(__file__))
     os.chdir("../../../../deployment/docker_compose/")
-    command = f"docker compose -f docker-compose.search-testing.yml -p danswer-stack{run_suffix} up -d"
+    command = f"docker compose -f docker-compose.search-testing.yml -p danswer-stack-{run_suffix} up -d"
     command += " --build"
     command += " --force-recreate"
-    if use_cloud_gpu:
-        command += " --scale indexing_model_server=0"
-        command += " --scale inference_model_server=0"
-    if launch_web_ui:
-        web_ui_port = 3000
-        while _is_port_in_use(web_ui_port):
-            web_ui_port += 1
-        print(f"UI will be launched at http://localhost:{web_ui_port}")
-        os.environ["NGINX_PORT"] = str(web_ui_port)
+
+    if only_state:
+        command += " index relational_db"
     else:
-        command += " --scale web_server=0"
-        command += " --scale nginx=0"
+        if use_cloud_gpu:
+            command += " --scale indexing_model_server=0"
+            command += " --scale inference_model_server=0"
+        if launch_web_ui:
+            web_ui_port = 3000
+            while _is_port_in_use(web_ui_port):
+                web_ui_port += 1
+            print(f"UI will be launched at http://localhost:{web_ui_port}")
+            os.environ["NGINX_PORT"] = str(web_ui_port)
+        else:
+            command += " --scale web_server=0"
+            command += " --scale nginx=0"
 
     print("Docker Command:\n", command)
 

backend/tests/regression/answer_quality/run_eval_pipeline.py

@@ -39,9 +39,11 @@ def main() -> None:
     if config.commit_sha:
         switch_to_commit(config.commit_sha)
 
-    start_docker_compose(run_suffix, config.launch_web_ui, config.use_cloud_gpu)
+    start_docker_compose(
+        run_suffix, config.launch_web_ui, config.use_cloud_gpu, config.only_state
+    )
 
-    if not config.existing_test_suffix:
+    if not config.existing_test_suffix and not config.only_state:
         upload_test_files(config.zipped_documents_file, run_suffix)
 
         run_qa_test_and_save_results(run_suffix)

backend/tests/regression/answer_quality/run_qa.py

@@ -46,19 +46,16 @@ def _get_test_output_folder(config: dict) -> str:
     base_output_folder = os.path.expanduser(config["output_folder"])
     if config["run_suffix"]:
         base_output_folder = os.path.join(
-            base_output_folder, ("test" + config["run_suffix"]), "evaluations_output"
+            base_output_folder, config["run_suffix"], "evaluations_output"
         )
     else:
         base_output_folder = os.path.join(base_output_folder, "no_defined_suffix")
 
     counter = 1
-    run_suffix = config["run_suffix"][1:]
-    output_folder_path = os.path.join(base_output_folder, f"{run_suffix}_run_1")
+    output_folder_path = os.path.join(base_output_folder, "run_1")
     while os.path.exists(output_folder_path):
         output_folder_path = os.path.join(
-            output_folder_path.replace(
-                f"{run_suffix}_run_{counter-1}", f"{run_suffix}_run_{counter}"
-            ),
+            output_folder_path.replace(f"run_{counter-1}", f"run_{counter}"),
         )
         counter += 1
 
@@ -163,6 +160,9 @@ def _process_and_write_query_results(config: dict) -> None:
         if not result.get("answer"):
             invalid_answer_count += 1
 
+        if not result.get("context_data_list"):
+            raise RuntimeError("Search failed, this is a critical failure!")
+
     _update_metadata_file(test_output_folder, invalid_answer_count)
 
     if invalid_answer_count:

backend/tests/regression/answer_quality/search_test_config.yaml.template

@@ -19,6 +19,9 @@ clean_up_docker_containers: true
 # Whether to launch a web UI for the test
 launch_web_ui: false
 
+# Whether to only run Vespa and Postgres
+only_state: false
+
 # Whether to use a cloud GPU for processing
 use_cloud_gpu: false
 

deployment/docker_compose/docker-compose.dev.yml

@@ -19,7 +19,7 @@ services:
     environment:
       # Auth Settings
       - AUTH_TYPE=${AUTH_TYPE:-disabled}
-      - SESSION_EXPIRE_TIME_SECONDS=${SESSION_EXPIRE_TIME_SECONDS:-86400}
+      - SESSION_EXPIRE_TIME_SECONDS=${SESSION_EXPIRE_TIME_SECONDS:-}
       - ENCRYPTION_KEY_SECRET=${ENCRYPTION_KEY_SECRET:-}
       - VALID_EMAIL_DOMAINS=${VALID_EMAIL_DOMAINS:-}
       - GOOGLE_OAUTH_CLIENT_ID=${GOOGLE_OAUTH_CLIENT_ID:-}

deployment/docker_compose/docker-compose.gpu-dev.yml

@@ -19,7 +19,7 @@ services:
     environment:
       # Auth Settings
       - AUTH_TYPE=${AUTH_TYPE:-disabled}
-      - SESSION_EXPIRE_TIME_SECONDS=${SESSION_EXPIRE_TIME_SECONDS:-86400}
+      - SESSION_EXPIRE_TIME_SECONDS=${SESSION_EXPIRE_TIME_SECONDS:-}
       - ENCRYPTION_KEY_SECRET=${ENCRYPTION_KEY_SECRET:-}
       - VALID_EMAIL_DOMAINS=${VALID_EMAIL_DOMAINS:-}
       - GOOGLE_OAUTH_CLIENT_ID=${GOOGLE_OAUTH_CLIENT_ID:-}

deployment/docker_compose/env.prod.template

@@ -57,8 +57,8 @@ SECRET=
 #SAML_CONF_DIR=
 
 
-# How long before user needs to reauthenticate, default to 1 day. (cookie expiration time)
-SESSION_EXPIRE_TIME_SECONDS=86400
+# How long before user needs to reauthenticate, default to 7 days. (cookie expiration time)
+SESSION_EXPIRE_TIME_SECONDS=604800
 
 
 # Use the below to specify a list of allowed user domains, only checked if user Auth is turned on

examples/widget/.eslintrc.json

@@ -0,0 +1,3 @@
+{
+  "extends": "next/core-web-vitals"
+}

examples/widget/.gitignore

@@ -0,0 +1,36 @@
+# See https://help.github.com/articles/ignoring-files/ for more about ignoring files.
+
+# dependencies
+/node_modules
+/.pnp
+.pnp.js
+.yarn/install-state.gz
+
+# testing
+/coverage
+
+# next.js
+/.next/
+/out/
+
+# production
+/build
+
+# misc
+.DS_Store
+*.pem
+
+# debug
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+
+# local env files
+.env*.local
+
+# vercel
+.vercel
+
+# typescript
+*.tsbuildinfo
+next-env.d.ts

examples/widget/README.md

@@ -0,0 +1,19 @@
+This is a code example for how you can use Danswer's APIs to build a chat bot widget for a website! The main code to look at can be found in `src/app/widget/Widget.tsx`.
+
+If you want to get fancier, then take a peek at the Chat implementation within Danswer itself [here](https://github.com/danswer-ai/danswer/blob/main/web/src/app/chat/ChatPage.tsx#L82).
+
+## Getting Started
+
+First, install the requirements:
+
+```bash
+npm i
+```
+
+Then run the development server:
+
+```bash
+npm run dev
+```
+
+Open [http://localhost:3000](http://localhost:3000) with your browser to see the result.

examples/widget/next.config.mjs

@@ -0,0 +1,4 @@
+/** @type {import('next').NextConfig} */
+const nextConfig = {};
+
+export default nextConfig;

examples/widget/package.json

@@ -0,0 +1,28 @@
+{
+  "name": "widget",
+  "version": "0.1.0",
+  "private": true,
+  "scripts": {
+    "dev": "next dev",
+    "build": "next build",
+    "start": "next start",
+    "lint": "next lint"
+  },
+  "dependencies": {
+    "next": "14.2.5",
+    "react": "^18",
+    "react-dom": "^18",
+    "react-markdown": "^8.0.6"
+  },
+  "devDependencies": {
+    "@types/node": "^20",
+    "@types/react": "^18",
+    "@types/react-dom": "^18",
+    "autoprefixer": "^10.4.19",
+    "eslint": "^8",
+    "eslint-config-next": "14.2.5",
+    "postcss": "^8.4.39",
+    "tailwindcss": "^3.4.6",
+    "typescript": "^5"
+  }
+}

examples/widget/postcss.config.mjs

@@ -0,0 +1,8 @@
+/** @type {import('postcss-load-config').Config} */
+const config = {
+  plugins: {
+    tailwindcss: {},
+  },
+};
+
+export default config;

examples/widget/src/app/globals.css

@@ -0,0 +1,3 @@
+@tailwind base;
+@tailwind components;
+@tailwind utilities;

examples/widget/src/app/layout.tsx

@@ -0,0 +1,23 @@
+import type { Metadata } from "next";
+import { Inter } from "next/font/google";
+
+import "./globals.css";
+
+const inter = Inter({ subsets: ["latin"] });
+
+export const metadata: Metadata = {
+  title: "Example Danswer Widget",
+  description: "Example Danswer Widget",
+};
+
+export default function RootLayout({
+  children,
+}: Readonly<{
+  children: React.ReactNode;
+}>) {
+  return (
+    <html lang="en">
+      <body className={inter.className}>{children}</body>
+    </html>
+  );
+}

examples/widget/src/app/page.tsx

@@ -0,0 +1,9 @@
+import { ChatWidget } from "./widget/Widget";
+
+export default function Home() {
+  return (
+    <main className="flex min-h-screen flex-col items-center justify-between p-24">
+      <ChatWidget />
+    </main>
+  );
+}

examples/widget/src/app/widget/Widget.tsx

@@ -0,0 +1,344 @@
+"use client";
+
+import React, { useState } from "react";
+import ReactMarkdown from "react-markdown";
+
+const API_URL = process.env.NEXT_PUBLIC_API_URL || "http://localhost:8080";
+const API_KEY = process.env.NEXT_PUBLIC_API_KEY || "";
+
+type NonEmptyObject = { [k: string]: any };
+
+const processSingleChunk = <T extends NonEmptyObject>(
+  chunk: string,
+  currPartialChunk: string | null,
+): [T | null, string | null] => {
+  const completeChunk = (currPartialChunk || "") + chunk;
+  try {
+    // every complete chunk should be valid JSON
+    const chunkJson = JSON.parse(completeChunk);
+    return [chunkJson, null];
+  } catch (err) {
+    // if it's not valid JSON, then it's probably an incomplete chunk
+    return [null, completeChunk];
+  }
+};
+
+const processRawChunkString = <T extends NonEmptyObject>(
+  rawChunkString: string,
+  previousPartialChunk: string | null,
+): [T[], string | null] => {
+  /* This is required because, in practice, we see that nginx does not send over
+  each chunk one at a time even with buffering turned off. Instead,
+  chunks are sometimes in batches or are sometimes incomplete */
+  if (!rawChunkString) {
+    return [[], null];
+  }
+  const chunkSections = rawChunkString
+    .split("\n")
+    .filter((chunk) => chunk.length > 0);
+  let parsedChunkSections: T[] = [];
+  let currPartialChunk = previousPartialChunk;
+  chunkSections.forEach((chunk) => {
+    const [processedChunk, partialChunk] = processSingleChunk<T>(
+      chunk,
+      currPartialChunk,
+    );
+    if (processedChunk) {
+      parsedChunkSections.push(processedChunk);
+      currPartialChunk = null;
+    } else {
+      currPartialChunk = partialChunk;
+    }
+  });
+
+  return [parsedChunkSections, currPartialChunk];
+};
+
+async function* handleStream<T extends NonEmptyObject>(
+  streamingResponse: Response,
+): AsyncGenerator<T[], void, unknown> {
+  const reader = streamingResponse.body?.getReader();
+  const decoder = new TextDecoder("utf-8");
+
+  let previousPartialChunk: string | null = null;
+  while (true) {
+    const rawChunk = await reader?.read();
+    if (!rawChunk) {
+      throw new Error("Unable to process chunk");
+    }
+    const { done, value } = rawChunk;
+    if (done) {
+      break;
+    }
+
+    const [completedChunks, partialChunk] = processRawChunkString<T>(
+      decoder.decode(value, { stream: true }),
+      previousPartialChunk,
+    );
+    if (!completedChunks.length && !partialChunk) {
+      break;
+    }
+    previousPartialChunk = partialChunk as string | null;
+
+    yield await Promise.resolve(completedChunks);
+  }
+}
+
+async function* sendMessage({
+  message,
+  chatSessionId,
+  parentMessageId,
+}: {
+  message: string;
+  chatSessionId?: number;
+  parentMessageId?: number;
+}) {
+  if (!chatSessionId || !parentMessageId) {
+    // Create a new chat session if one doesn't exist
+    const createSessionResponse = await fetch(
+      `${API_URL}/chat/create-chat-session`,
+      {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          Authorization: `Bearer ${API_KEY}`,
+        },
+        body: JSON.stringify({
+          // or specify an assistant you have defined
+          persona_id: 0,
+        }),
+      },
+    );
+
+    if (!createSessionResponse.ok) {
+      const errorJson = await createSessionResponse.json();
+      const errorMsg = errorJson.message || errorJson.detail || "";
+      throw Error(`Failed to create chat session - ${errorMsg}`);
+    }
+
+    const sessionData = await createSessionResponse.json();
+    chatSessionId = sessionData.chat_session_id;
+  }
+
+  const sendMessageResponse = await fetch(`${API_URL}/chat/send-message`, {
+    method: "POST",
+    headers: {
+      "Content-Type": "application/json",
+      Authorization: `Bearer ${API_KEY}`,
+    },
+    body: JSON.stringify({
+      chat_session_id: chatSessionId,
+      parent_message_id: parentMessageId,
+      message: message,
+      prompt_id: null,
+      search_doc_ids: null,
+      file_descriptors: [],
+      // checkout https://github.com/danswer-ai/danswer/blob/main/backend/danswer/search/models.py#L105 for
+      // all available options
+      retrieval_options: {
+        run_search: "always",
+        filters: null,
+      },
+      query_override: null,
+    }),
+  });
+  if (!sendMessageResponse.ok) {
+    const errorJson = await sendMessageResponse.json();
+    const errorMsg = errorJson.message || errorJson.detail || "";
+    throw Error(`Failed to send message - ${errorMsg}`);
+  }
+
+  yield* handleStream<NonEmptyObject>(sendMessageResponse);
+}
+
+export const ChatWidget = () => {
+  const [messages, setMessages] = useState<{ text: string; isUser: boolean }[]>(
+    [],
+  );
+  const [inputText, setInputText] = useState("");
+  const [isLoading, setIsLoading] = useState(false);
+
+  const handleSubmit = async (e: React.FormEvent) => {
+    e.preventDefault();
+    if (inputText.trim()) {
+      const initialPrevMessages = messages;
+      setMessages([...initialPrevMessages, { text: inputText, isUser: true }]);
+      setInputText("");
+      setIsLoading(true);
+
+      try {
+        const messageGenerator = sendMessage({
+          message: inputText,
+          chatSessionId: undefined,
+          parentMessageId: undefined,
+        });
+        let fullResponse = "";
+
+        for await (const chunks of messageGenerator) {
+          for (const chunk of chunks) {
+            if ("answer_piece" in chunk) {
+              fullResponse += chunk.answer_piece;
+              setMessages([
+                ...initialPrevMessages,
+                { text: inputText, isUser: true },
+                { text: fullResponse, isUser: false },
+              ]);
+            }
+          }
+        }
+      } catch (error) {
+        console.error("Error sending message:", error);
+        setMessages((prevMessages) => [
+          ...prevMessages,
+          { text: "An error occurred. Please try again.", isUser: false },
+        ]);
+      } finally {
+        setIsLoading(false);
+      }
+    }
+  };
+
+  return (
+    <div
+      className="
+      fixed
+      bottom-4
+      right-4
+      z-50
+      bg-white
+      rounded-lg
+      shadow-xl
+      w-96
+      h-[32rem]
+      flex
+      flex-col
+      overflow-hidden
+      transition-all
+      duration-300
+      ease-in-out
+    "
+    >
+      <div
+        className="
+        bg-gradient-to-r
+        from-blue-600
+        to-blue-800
+        text-white
+        p-4
+        font-bold
+        flex
+        justify-between
+        items-center
+      "
+      >
+        <span>Chat Support</span>
+      </div>
+      <div
+        className="
+        flex-grow
+        overflow-y-auto
+        p-4
+        space-y-4
+        bg-gray-50
+        border-b
+        border-gray-200
+      "
+      >
+        {messages.map((message, index) => (
+          <div
+            key={index}
+            className={`
+            flex
+            ${message.isUser ? "justify-end" : "justify-start"}
+          `}
+          >
+            <div
+              className={`
+              max-w-[75%]
+              p-3
+              rounded-lg
+              ${
+                message.isUser
+                  ? "bg-blue-500 text-white"
+                  : "bg-white text-black"
+              }
+              shadow
+            `}
+            >
+              <ReactMarkdown>{message.text}</ReactMarkdown>
+            </div>
+          </div>
+        ))}
+        {isLoading && (
+          <div className="flex justify-center">
+            <div className="animate-pulse flex space-x-2">
+              <div className="w-2 h-2 bg-gray-500 rounded-full"></div>
+              <div className="w-2 h-2 bg-gray-500 rounded-full"></div>
+              <div className="w-2 h-2 bg-gray-500 rounded-full"></div>
+            </div>
+          </div>
+        )}
+      </div>
+      <form
+        onSubmit={handleSubmit}
+        className="
+        p-4
+        bg-white
+        border-t
+        border-gray-200
+      "
+      >
+        <div className="relative">
+          <input
+            type="text"
+            value={inputText}
+            onChange={(e) => setInputText(e.target.value)}
+            placeholder="Type a message..."
+            className="
+              w-full
+              p-2
+              pr-10
+              border
+              border-gray-300
+              rounded-full
+              focus:outline-none
+              focus:ring-2
+              focus:ring-blue-500
+              focus:border-transparent
+            "
+            disabled={isLoading}
+          />
+          <button
+            type="submit"
+            disabled={isLoading}
+            className="
+              absolute
+              right-2
+              top-1/2
+              transform
+              -translate-y-1/2
+              text-blue-500
+              hover:text-blue-600
+              focus:outline-none
+            "
+          >
+            <svg
+              xmlns="http://www.w3.org/2000/svg"
+              className="h-6 w-6"
+              fill="none"
+              viewBox="0 0 24 24"
+              stroke="currentColor"
+            >
+              <path
+                strokeLinecap="round"
+                strokeLinejoin="round"
+                strokeWidth={2}
+                d="M12 19l9 2-9-18-9 18 9-2zm0 0v-8"
+              />
+            </svg>
+          </button>
+        </div>
+      </form>
+    </div>
+  );
+};

examples/widget/tailwind.config.ts

@@ -0,0 +1,20 @@
+import type { Config } from "tailwindcss";
+
+const config: Config = {
+  content: [
+    "./src/pages/**/*.{js,ts,jsx,tsx,mdx}",
+    "./src/components/**/*.{js,ts,jsx,tsx,mdx}",
+    "./src/app/**/*.{js,ts,jsx,tsx,mdx}",
+  ],
+  theme: {
+    extend: {
+      backgroundImage: {
+        "gradient-radial": "radial-gradient(var(--tw-gradient-stops))",
+        "gradient-conic":
+          "conic-gradient(from 180deg at 50% 50%, var(--tw-gradient-stops))",
+      },
+    },
+  },
+  plugins: [],
+};
+export default config;

examples/widget/tsconfig.json

@@ -0,0 +1,26 @@
+{
+  "compilerOptions": {
+    "lib": ["dom", "dom.iterable", "esnext"],
+    "allowJs": true,
+    "skipLibCheck": true,
+    "strict": true,
+    "noEmit": true,
+    "esModuleInterop": true,
+    "module": "esnext",
+    "moduleResolution": "bundler",
+    "resolveJsonModule": true,
+    "isolatedModules": true,
+    "jsx": "preserve",
+    "incremental": true,
+    "plugins": [
+      {
+        "name": "next"
+      }
+    ],
+    "paths": {
+      "@/*": ["./src/*"]
+    }
+  },
+  "include": ["next-env.d.ts", "**/*.ts", "**/*.tsx", ".next/types/**/*.ts"],
+  "exclude": ["node_modules"]
+}

web/src/app/assistants/SidebarWrapper.tsx

@@ -9,7 +9,7 @@ import { Folder } from "@/app/chat/folders/interfaces";
 import { User } from "@/lib/types";
 import { Persona } from "@/app/admin/assistants/interfaces";
 import Cookies from "js-cookie";
-import { SIDEBAR_TOGGLED_COOKIE_NAME } from "@/components/resizable/contants";
+import { SIDEBAR_TOGGLED_COOKIE_NAME } from "@/components/resizable/constants";
 import { ReactNode, useEffect, useRef, useState } from "react";
 import { useSidebarVisibility } from "@/components/chat_search/hooks";
 import FunctionalHeader from "@/components/chat_search/Header";
@@ -143,6 +143,7 @@ export default function SidebarWrapper<T extends object>({
           <div className="mt-4 mx-auto">{content(contentProps)}</div>
         </div>
       </div>
+      <FixedLogo />
     </div>
   );
 }

web/src/app/assistants/gallery/page.tsx

@@ -65,9 +65,6 @@ export default async function GalleryPage({
           user={user}
           assistants={assistants}
         />
-
-        {/* Temporary - fixed logo */}
-        <FixedLogo />
       </ChatProvider>
     </>
   );

web/src/app/assistants/mine/page.tsx

@@ -67,9 +67,6 @@ export default async function GalleryPage({
           user={user}
           assistants={assistants}
         />
-
-        {/* Temporary - fixed logo */}
-        <FixedLogo />
       </ChatProvider>
     </>
   );

web/src/app/chat/ChatPage.tsx

@@ -72,7 +72,7 @@ import { ChatBanner } from "./ChatBanner";
 
 import FunctionalHeader from "@/components/chat_search/Header";
 import { useSidebarVisibility } from "@/components/chat_search/hooks";
-import { SIDEBAR_TOGGLED_COOKIE_NAME } from "@/components/resizable/contants";
+import { SIDEBAR_TOGGLED_COOKIE_NAME } from "@/components/resizable/constants";
 import FixedLogo from "./shared_chat_search/FixedLogo";
 
 const TEMP_USER_MESSAGE_ID = -1;

web/src/app/chat/message/Messages.tsx

@@ -57,6 +57,8 @@ import {
 import { ValidSources } from "@/lib/types";
 import { Tooltip } from "@/components/tooltip/Tooltip";
 import { useMouseTracking } from "./hooks";
+import { InternetSearchIcon } from "@/components/InternetSearchIcon";
+import { getTitleFromDocument } from "@/lib/sources";
 
 const TOOLS_WITH_CUSTOM_HANDLING = [
   SEARCH_TOOL_NAME,
@@ -443,17 +445,17 @@ export const AIMessage = ({
                                   >
                                     <Citation link={doc.link} index={ind + 1} />
                                     <p className="shrink truncate ellipsis break-all ">
-                                      {
-                                        doc.document_id.split("/")[
-                                          doc.document_id.split("/").length - 1
-                                        ]
-                                      }
+                                      {getTitleFromDocument(doc)}
                                     </p>
                                     <div className="ml-auto flex-none">
-                                      <SourceIcon
-                                        sourceType={doc.source_type}
-                                        iconSize={18}
-                                      />
+                                      {doc.is_internet ? (
+                                        <InternetSearchIcon url={doc.link} />
+                                      ) : (
+                                        <SourceIcon
+                                          sourceType={doc.source_type}
+                                          iconSize={18}
+                                        />
+                                      )}
                                     </div>
                                   </a>
                                   <div className="flex overscroll-x-scroll mt-.5">

web/src/app/chat/sessionSidebar/PagesTab.tsx

@@ -54,11 +54,13 @@ export function PagesTab({
     }
   };
 
+  const isHistoryEmpty = !existingChats || existingChats.length === 0;
+
   return (
     <div className="mb-1 ml-3 relative miniscroll overflow-y-auto h-full">
       {folders && folders.length > 0 && (
         <div className="py-2 border-b border-border">
-          <div className="text-xs text-subtle flex pb-0.5 mb-1.5 mt-2 font-medium">
+          <div className="text-xs text-subtle flex pb-0.5 mb-1.5 mt-2 font-bold">
             Folders
           </div>
           <FolderList
@@ -80,43 +82,51 @@ export function PagesTab({
         } rounded-md`}
       >
         {(page == "chat" || page == "search") && (
-          <p className="my-2 text-xs text-subtle flex font-semibold">
+          <p className="my-2 text-xs text-subtle flex font-bold">
             {page == "chat" && "Chat "}
             {page == "search" && "Search "}
             History
           </p>
         )}
-        {Object.entries(groupedChatSessions).map(
-          ([dateRange, chatSessions], ind) => {
-            if (chatSessions.length > 0) {
-              return (
-                <div key={dateRange}>
-                  <div
-                    className={`text-xs text-subtle  ${
-                      ind != 0 && "mt-5"
-                    } flex pb-0.5 mb-1.5 font-medium`}
-                  >
-                    {dateRange}
+        {isHistoryEmpty ? (
+          <p className="text-sm text-subtle mt-2 w-[250px]">
+            {page === "search"
+              ? "Try running a search! Your search history will appear here."
+              : "Try sending a message! Your chat history will appear here."}
+          </p>
+        ) : (
+          Object.entries(groupedChatSessions).map(
+            ([dateRange, chatSessions], ind) => {
+              if (chatSessions.length > 0) {
+                return (
+                  <div key={dateRange}>
+                    <div
+                      className={`text-xs text-subtle ${
+                        ind != 0 && "mt-5"
+                      } flex pb-0.5 mb-1.5 font-medium`}
+                    >
+                      {dateRange}
+                    </div>
+                    {chatSessions
+                      .filter((chat) => chat.folder_id === null)
+                      .map((chat) => {
+                        const isSelected = currentChatId === chat.id;
+                        return (
+                          <div key={`${chat.id}-${chat.name}`}>
+                            <ChatSessionDisplay
+                              search={page == "search"}
+                              chatSession={chat}
+                              isSelected={isSelected}
+                              skipGradient={isDragOver}
+                            />
+                          </div>
+                        );
+                      })}
                   </div>
-                  {chatSessions
-                    .filter((chat) => chat.folder_id === null)
-                    .map((chat) => {
-                      const isSelected = currentChatId === chat.id;
-                      return (
-                        <div key={`${chat.id}-${chat.name}`}>
-                          <ChatSessionDisplay
-                            search={page == "search"}
-                            chatSession={chat}
-                            isSelected={isSelected}
-                            skipGradient={isDragOver}
-                          />
-                        </div>
-                      );
-                    })}
-                </div>
-              );
+                );
+              }
             }
-          }
+          )
         )}
       </div>
     </div>

web/src/app/chat/shared_chat_search/FixedLogo.tsx

@@ -1,3 +1,5 @@
+"use client";
+
 import { HeaderTitle } from "@/components/header/Header";
 import { Logo } from "@/components/Logo";
 import { SettingsContext } from "@/components/settings/SettingsProvider";

web/src/app/search/page.tsx

@@ -30,7 +30,7 @@ import {
 } from "@/lib/assistants/fetchAssistantsSS";
 import FunctionalWrapper from "../chat/shared_chat_search/FunctionalWrapper";
 import { ChatSession } from "../chat/interfaces";
-import { SIDEBAR_TOGGLED_COOKIE_NAME } from "@/components/resizable/contants";
+import { SIDEBAR_TOGGLED_COOKIE_NAME } from "@/components/resizable/constants";
 import ToggleSearch from "./WrappedSearch";
 import {
   AGENTIC_SEARCH_TYPE_COOKIE_NAME,

web/src/components/admin/Layout.tsx

@@ -329,17 +329,10 @@ export async function Layout({ children }: { children: React.ReactNode }) {
           />
         </div>
         <div className="pb-8 relative h-full overflow-y-auto w-full">
-          <div className="fixed bg-background left-0 border-b gap-x-4 mb-8 px-4 py-2 w-full items-center flex justify-end">
-            <a
-              href="/chat"
-              className="transition-all duration-150 cursor-pointer p-1 text-sm items-center flex gap-x-1 px-2 py-1 rounded-lg hover:shadow-sm hover:ring-1 hover:ring-ingio-900/40 hover:bg-opacity-90 text-neutral-100 bg-accent"
-            >
-              <BackIcon size={20} className="text-neutral" />
-              Back to Danswer
-            </a>
+          <div className="fixed bg-background left-0 gap-x-4 mb-8 px-4 py-2 w-full items-center flex justify-end">
             <UserDropdown user={user} />
           </div>
-          <div className="pt-20 flex overflow-y-auto h-full px-4 md:px-12">
+          <div className="pt-12 flex overflow-y-auto h-full px-4 md:px-12">
             {children}
           </div>
         </div>

web/src/components/admin/connectors/AdminSidebar.tsx

@@ -6,6 +6,7 @@ import { Logo } from "@/components/Logo";
 import { NEXT_PUBLIC_DO_NOT_USE_TOGGLE_OFF_DANSWER_POWERED } from "@/lib/constants";
 import { HeaderTitle } from "@/components/header/Header";
 import { SettingsContext } from "@/components/settings/SettingsProvider";
+import { BackIcon } from "@/components/icons/icons";
 
 interface Item {
   name: string | JSX.Element;
@@ -62,6 +63,12 @@ export function AdminSidebar({ collections }: { collections: Collection[] }) {
             </Link>
           </div>
         </div>
+        <Link href={"/chat"}>
+          <button className="text-sm block w-48 py-2.5 flex px-2 text-left bg-background-200 hover:bg-background-200/80 cursor-pointer rounded">
+            <BackIcon size={20} className="text-neutral" />
+            <p className="ml-1">Back to Danswer</p>
+          </button>
+        </Link>
         {collections.map((collection, collectionInd) => (
           <div key={collectionInd}>
             <h2 className="text-xs text-strong font-bold pb-2">
@@ -70,7 +77,7 @@ export function AdminSidebar({ collections }: { collections: Collection[] }) {
             {collection.items.map((item) => (
               <Link key={item.link} href={item.link}>
                 <button className="text-sm block w-48 py-2.5 px-2 text-left hover:bg-hover rounded">
-                  <div className="">{item.name}</div>
+                  {item.name}
                 </button>
               </Link>
             ))}

web/src/components/resizable/ResizableSection.tsx

@@ -2,7 +2,7 @@
 
 import React, { useEffect, useState } from "react";
 import Cookies from "js-cookie";
-import { DOCUMENT_SIDEBAR_WIDTH_COOKIE_NAME } from "./contants";
+import { DOCUMENT_SIDEBAR_WIDTH_COOKIE_NAME } from "./constants";
 
 function applyMinAndMax(
   width: number,

web/src/components/search/SearchBar.tsx

@@ -191,6 +191,12 @@ export const FullSearchBar = ({
 
       <div className="flex justify-end w-full items-center space-x-3 mr-12 px-4 pb-2">
         {searchState == "searching" && (
+          <div key={"Reading"} className="mr-auto relative inline-block">
+            <span className="loading-text">Searching...</span>
+          </div>
+        )}
+
+        {searchState == "reading" && (
           <div key={"Reading"} className="mr-auto relative inline-block">
             <span className="loading-text">Reading Documents...</span>
           </div>

web/src/components/search/SearchSection.tsx

@@ -29,12 +29,17 @@ import { HistorySidebar } from "@/app/chat/sessionSidebar/HistorySidebar";
 import { ChatSession, SearchSession } from "@/app/chat/interfaces";
 import FunctionalHeader from "../chat_search/Header";
 import { useSidebarVisibility } from "../chat_search/hooks";
-import { SIDEBAR_TOGGLED_COOKIE_NAME } from "../resizable/contants";
+import { SIDEBAR_TOGGLED_COOKIE_NAME } from "../resizable/constants";
 import { AGENTIC_SEARCH_TYPE_COOKIE_NAME } from "@/lib/constants";
 import Cookies from "js-cookie";
 import FixedLogo from "@/app/chat/shared_chat_search/FixedLogo";
 
-export type searchState = "input" | "searching" | "analyzing" | "summarizing";
+export type searchState =
+  | "input"
+  | "searching"
+  | "reading"
+  | "analyzing"
+  | "summarizing";
 
 const SEARCH_DEFAULT_OVERRIDES_START: SearchDefaultOverrides = {
   forceDisplayQA: false,
@@ -231,10 +236,23 @@ export const SearchSection = ({
 
   const updateDocs = (documents: SearchDanswerDocument[]) => {
     setTimeout(() => {
-      if (searchState != "input") {
-        setSearchState("analyzing");
-      }
+      setSearchState((searchState) => {
+        if (searchState != "input") {
+          return "reading";
+        }
+        return "input";
+      });
     }, 1500);
+
+    setTimeout(() => {
+      setSearchState((searchState) => {
+        if (searchState != "input") {
+          return "analyzing";
+        }
+        return "input";
+      });
+    }, 4500);
+
     setSearchResponse((prevState) => ({
       ...(prevState || initialSearchResponse),
       documents,

web/src/lib/chat/fetchChatData.ts

@@ -23,7 +23,7 @@ import { cookies } from "next/headers";
 import {
   SIDEBAR_TOGGLED_COOKIE_NAME,
   DOCUMENT_SIDEBAR_WIDTH_COOKIE_NAME,
-} from "@/components/resizable/contants";
+} from "@/components/resizable/constants";
 import { hasCompletedWelcomeFlowSS } from "@/components/initialSetup/welcome/WelcomeModalWrapper";
 import { fetchAssistantsSS } from "../assistants/fetchAssistantsSS";
 

web/src/lib/llm/utils.ts

@@ -36,10 +36,12 @@ export function getFinalLLM(
 
 const MODELS_SUPPORTING_IMAGES = [
   ["openai", "gpt-4o"],
+  ["openai", "gpt-4o-mini"],
   ["openai", "gpt-4-vision-preview"],
   ["openai", "gpt-4-turbo"],
   ["openai", "gpt-4-1106-vision-preview"],
   ["azure", "gpt-4o"],
+  ["azure", "gpt-4o-mini"],
   ["azure", "gpt-4-vision-preview"],
   ["azure", "gpt-4-turbo"],
   ["azure", "gpt-4-1106-vision-preview"],

web/src/lib/search/streamingQa.ts

@@ -61,7 +61,7 @@ export const searchRequestStreamed = async ({
           filters: filters,
           enable_auto_detect_filters: false,
         },
-        evaluate_response: true,
+        llm_doc_eval: true,
         skip_gen_ai_answer_generation: true,
       }),
       headers: {

web/src/lib/sources.ts

@@ -38,7 +38,11 @@ import {
   ColorSlackIcon,
 } from "@/components/icons/icons";
 import { ValidSources } from "./types";
-import { SourceCategory, SourceMetadata } from "./search/interfaces";
+import {
+  DanswerDocument,
+  SourceCategory,
+  SourceMetadata,
+} from "./search/interfaces";
 import { Persona } from "@/app/admin/assistants/interfaces";
 import { FaAccessibleIcon, FaSlack } from "react-icons/fa";
 
@@ -290,3 +294,16 @@ export function getSourcesForPersona(persona: Persona): ValidSources[] {
   });
   return personaSources;
 }
+
+function stripTrailingSlash(str: string) {
+  if (str.substr(-1) === "/") {
+    return str.substr(0, str.length - 1);
+  }
+  return str;
+}
+
+export const getTitleFromDocument = (document: DanswerDocument) => {
+  return stripTrailingSlash(document.document_id).split("/")[
+    stripTrailingSlash(document.document_id).split("/").length - 1
+  ];
+};