updates

refactor

.github/workflows/pr-python-connector-tests.yml

@@ -24,6 +24,8 @@ env:
   GOOGLE_DRIVE_OAUTH_CREDENTIALS_JSON_STR: ${{ secrets.GOOGLE_DRIVE_OAUTH_CREDENTIALS_JSON_STR }}
   GOOGLE_GMAIL_SERVICE_ACCOUNT_JSON_STR: ${{ secrets.GOOGLE_GMAIL_SERVICE_ACCOUNT_JSON_STR }}
   GOOGLE_GMAIL_OAUTH_CREDENTIALS_JSON_STR: ${{ secrets.GOOGLE_GMAIL_OAUTH_CREDENTIALS_JSON_STR }}
+  # Slab
+  SLAB_BOT_TOKEN: ${{ secrets.SLAB_BOT_TOKEN }}
 
 jobs:
   connectors-check:

backend/Dockerfile

@@ -73,6 +73,7 @@ RUN apt-get update && \
     rm -rf /var/lib/apt/lists/* && \
     rm -f /usr/local/lib/python3.11/site-packages/tornado/test/test.key
 
+
 # Pre-downloading models for setups with limited egress
 RUN python -c "from tokenizers import Tokenizer; \
 Tokenizer.from_pretrained('nomic-ai/nomic-embed-text-v1')"

backend/alembic/env.py

@@ -1,5 +1,5 @@
 from sqlalchemy.engine.base import Connection
-from typing import Any
+from typing import Literal
 import asyncio
 from logging.config import fileConfig
 import logging
@@ -8,6 +8,7 @@ from alembic import context
 from sqlalchemy import pool
 from sqlalchemy.ext.asyncio import create_async_engine
 from sqlalchemy.sql import text
+from sqlalchemy.sql.schema import SchemaItem
 
 from shared_configs.configs import MULTI_TENANT
 from danswer.db.engine import build_connection_string
@@ -35,7 +36,18 @@ logger = logging.getLogger(__name__)
 
 
 def include_object(
-    object: Any, name: str, type_: str, reflected: bool, compare_to: Any
+    object: SchemaItem,
+    name: str | None,
+    type_: Literal[
+        "schema",
+        "table",
+        "column",
+        "index",
+        "unique_constraint",
+        "foreign_key_constraint",
+    ],
+    reflected: bool,
+    compare_to: SchemaItem | None,
 ) -> bool:
     """
     Determines whether a database object should be included in migrations.

backend/alembic/versions/93560ba1b118_add_web_ui_option_to_slack_config.py

@@ -0,0 +1,35 @@
+"""add web ui option to slack config
+
+Revision ID: 93560ba1b118
+Revises: 6d562f86c78b
+Create Date: 2024-11-24 06:36:17.490612
+
+"""
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = "93560ba1b118"
+down_revision = "6d562f86c78b"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # Add show_continue_in_web_ui with default False to all existing channel_configs
+    op.execute(
+        """
+        UPDATE slack_channel_config
+        SET channel_config = channel_config || '{"show_continue_in_web_ui": false}'::jsonb
+        WHERE NOT channel_config ? 'show_continue_in_web_ui'
+        """
+    )
+
+
+def downgrade() -> None:
+    # Remove show_continue_in_web_ui from all channel_configs
+    op.execute(
+        """
+        UPDATE slack_channel_config
+        SET channel_config = channel_config - 'show_continue_in_web_ui'
+        """
+    )

backend/alembic/versions/9f696734098f_combine_search_and_chat.py

@@ -0,0 +1,36 @@
+"""Combine Search and Chat
+
+Revision ID: 9f696734098f
+Revises: a8c2065484e6
+Create Date: 2024-11-27 15:32:19.694972
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "9f696734098f"
+down_revision = "a8c2065484e6"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.alter_column("chat_session", "description", nullable=True)
+    op.drop_column("chat_session", "one_shot")
+    op.drop_column("slack_channel_config", "response_type")
+
+
+def downgrade() -> None:
+    op.execute("UPDATE chat_session SET description = '' WHERE description IS NULL")
+    op.alter_column("chat_session", "description", nullable=False)
+    op.add_column(
+        "chat_session",
+        sa.Column("one_shot", sa.Boolean(), nullable=False, server_default=sa.false()),
+    )
+    op.add_column(
+        "slack_channel_config",
+        sa.Column(
+            "response_type", sa.String(), nullable=False, server_default="citations"
+        ),
+    )

backend/alembic/versions/a8c2065484e6_add_auto_scroll_to_user_model.py

@@ -0,0 +1,27 @@
+"""add auto scroll to user model
+
+Revision ID: a8c2065484e6
+Revises: abe7378b8217
+Create Date: 2024-11-22 17:34:09.690295
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "a8c2065484e6"
+down_revision = "abe7378b8217"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "user",
+        sa.Column("auto_scroll", sa.Boolean(), nullable=True, server_default=None),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("user", "auto_scroll")

backend/alembic/versions/abe7378b8217_add_indexing_trigger_to_cc_pair.py

@@ -0,0 +1,30 @@
+"""add indexing trigger to cc_pair
+
+Revision ID: abe7378b8217
+Revises: 6d562f86c78b
+Create Date: 2024-11-26 19:09:53.481171
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "abe7378b8217"
+down_revision = "93560ba1b118"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "connector_credential_pair",
+        sa.Column(
+            "indexing_trigger",
+            sa.Enum("UPDATE", "REINDEX", name="indexingmode", native_enum=False),
+            nullable=True,
+        ),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("connector_credential_pair", "indexing_trigger")

backend/alembic_tenants/env.py

@@ -1,5 +1,6 @@
 import asyncio
 from logging.config import fileConfig
+from typing import Literal
 
 from sqlalchemy import pool
 from sqlalchemy.engine import Connection
@@ -37,8 +38,15 @@ EXCLUDE_TABLES = {"kombu_queue", "kombu_message"}
 
 def include_object(
     object: SchemaItem,
-    name: str,
-    type_: str,
+    name: str | None,
+    type_: Literal[
+        "schema",
+        "table",
+        "column",
+        "index",
+        "unique_constraint",
+        "foreign_key_constraint",
+    ],
     reflected: bool,
     compare_to: SchemaItem | None,
 ) -> bool:

backend/danswer/access/models.py

@@ -18,6 +18,11 @@ class ExternalAccess:
 
 @dataclass(frozen=True)
 class DocExternalAccess:
+    """
+    This is just a class to wrap the external access and the document ID
+    together. It's used for syncing document permissions to Redis.
+    """
+
     external_access: ExternalAccess
     # The document ID
     doc_id: str

backend/danswer/agent_search/answer_query/graph_builder.py

@@ -0,0 +1,100 @@
+from langgraph.graph import END
+from langgraph.graph import START
+from langgraph.graph import StateGraph
+
+from danswer.agent_search.answer_query.nodes.answer_check import answer_check
+from danswer.agent_search.answer_query.nodes.answer_generation import answer_generation
+from danswer.agent_search.answer_query.nodes.format_answer import format_answer
+from danswer.agent_search.answer_query.states import AnswerQueryInput
+from danswer.agent_search.answer_query.states import AnswerQueryOutput
+from danswer.agent_search.answer_query.states import AnswerQueryState
+from danswer.agent_search.expanded_retrieval.graph_builder import (
+    expanded_retrieval_graph_builder,
+)
+
+
+def answer_query_graph_builder() -> StateGraph:
+    graph = StateGraph(
+        state_schema=AnswerQueryState,
+        input=AnswerQueryInput,
+        output=AnswerQueryOutput,
+    )
+
+    ### Add nodes ###
+
+    expanded_retrieval = expanded_retrieval_graph_builder().compile()
+    graph.add_node(
+        node="expanded_retrieval_for_initial_decomp",
+        action=expanded_retrieval,
+    )
+    graph.add_node(
+        node="answer_check",
+        action=answer_check,
+    )
+    graph.add_node(
+        node="answer_generation",
+        action=answer_generation,
+    )
+    graph.add_node(
+        node="format_answer",
+        action=format_answer,
+    )
+
+    ### Add edges ###
+
+    graph.add_edge(
+        start_key=START,
+        end_key="expanded_retrieval_for_initial_decomp",
+    )
+    graph.add_edge(
+        start_key="expanded_retrieval_for_initial_decomp",
+        end_key="answer_generation",
+    )
+    graph.add_edge(
+        start_key="answer_generation",
+        end_key="answer_check",
+    )
+    graph.add_edge(
+        start_key="answer_check",
+        end_key="format_answer",
+    )
+    graph.add_edge(
+        start_key="format_answer",
+        end_key=END,
+    )
+
+    return graph
+
+
+if __name__ == "__main__":
+    from danswer.db.engine import get_session_context_manager
+    from danswer.llm.factory import get_default_llms
+    from danswer.context.search.models import SearchRequest
+
+    graph = answer_query_graph_builder()
+    compiled_graph = graph.compile()
+    primary_llm, fast_llm = get_default_llms()
+    search_request = SearchRequest(
+        query="Who made Excel and what other products did they make?",
+    )
+    with get_session_context_manager() as db_session:
+        inputs = AnswerQueryInput(
+            search_request=search_request,
+            primary_llm=primary_llm,
+            fast_llm=fast_llm,
+            db_session=db_session,
+            query_to_answer="Who made Excel?",
+        )
+        output = compiled_graph.invoke(
+            input=inputs,
+            # debug=True,
+            # subgraphs=True,
+        )
+        print(output)
+        # for namespace, chunk in compiled_graph.stream(
+        #     input=inputs,
+        #     # debug=True,
+        #     subgraphs=True,
+        # ):
+        #     print(namespace)
+        #     print(chunk)

backend/danswer/agent_search/answer_query/nodes/answer_check.py

@@ -0,0 +1,30 @@
+from langchain_core.messages import HumanMessage
+from langchain_core.messages import merge_message_runs
+
+from danswer.agent_search.answer_query.states import AnswerQueryState
+from danswer.agent_search.answer_query.states import QACheckOutput
+from danswer.agent_search.shared_graph_utils.prompts import BASE_CHECK_PROMPT
+
+
+def answer_check(state: AnswerQueryState) -> QACheckOutput:
+    msg = [
+        HumanMessage(
+            content=BASE_CHECK_PROMPT.format(
+                question=state["search_request"].query,
+                base_answer=state["answer"],
+            )
+        )
+    ]
+
+    fast_llm = state["fast_llm"]
+    response = list(
+        fast_llm.stream(
+            prompt=msg,
+        )
+    )
+
+    response_str = merge_message_runs(response, chunk_separator="")[0].content
+
+    return QACheckOutput(
+        answer_quality=response_str,
+    )

backend/danswer/agent_search/answer_query/nodes/answer_generation.py

@@ -0,0 +1,32 @@
+from langchain_core.messages import HumanMessage
+from langchain_core.messages import merge_message_runs
+
+from danswer.agent_search.answer_query.states import AnswerQueryState
+from danswer.agent_search.answer_query.states import QAGenerationOutput
+from danswer.agent_search.shared_graph_utils.prompts import BASE_RAG_PROMPT
+from danswer.agent_search.shared_graph_utils.utils import format_docs
+
+
+def answer_generation(state: AnswerQueryState) -> QAGenerationOutput:
+    query = state["query_to_answer"]
+    docs = state["reordered_documents"]
+
+    print(f"Number of verified retrieval docs: {len(docs)}")
+
+    msg = [
+        HumanMessage(
+            content=BASE_RAG_PROMPT.format(question=query, context=format_docs(docs))
+        )
+    ]
+
+    fast_llm = state["fast_llm"]
+    response = list(
+        fast_llm.stream(
+            prompt=msg,
+        )
+    )
+
+    answer_str = merge_message_runs(response, chunk_separator="")[0].content
+    return QAGenerationOutput(
+        answer=answer_str,
+    )

backend/danswer/agent_search/answer_query/nodes/format_answer.py

@@ -0,0 +1,16 @@
+from danswer.agent_search.answer_query.states import AnswerQueryOutput
+from danswer.agent_search.answer_query.states import AnswerQueryState
+from danswer.agent_search.answer_query.states import SearchAnswerResults
+
+
+def format_answer(state: AnswerQueryState) -> AnswerQueryOutput:
+    return AnswerQueryOutput(
+        decomp_answer_results=[
+            SearchAnswerResults(
+                query=state["query_to_answer"],
+                quality=state["answer_quality"],
+                answer=state["answer"],
+                documents=state["reordered_documents"],
+            )
+        ],
+    )

backend/danswer/agent_search/answer_query/states.py

@@ -0,0 +1,45 @@
+from typing import Annotated
+from typing import TypedDict
+
+from pydantic import BaseModel
+
+from danswer.agent_search.core_state import PrimaryState
+from danswer.agent_search.shared_graph_utils.operators import dedup_inference_sections
+from danswer.context.search.models import InferenceSection
+
+
+class SearchAnswerResults(BaseModel):
+    query: str
+    answer: str
+    quality: str
+    documents: Annotated[list[InferenceSection], dedup_inference_sections]
+
+
+class QACheckOutput(TypedDict, total=False):
+    answer_quality: str
+
+
+class QAGenerationOutput(TypedDict, total=False):
+    answer: str
+
+
+class ExpandedRetrievalOutput(TypedDict):
+    reordered_documents: Annotated[list[InferenceSection], dedup_inference_sections]
+
+
+class AnswerQueryState(
+    PrimaryState,
+    QACheckOutput,
+    QAGenerationOutput,
+    ExpandedRetrievalOutput,
+    total=True,
+):
+    query_to_answer: str
+
+
+class AnswerQueryInput(PrimaryState, total=True):
+    query_to_answer: str
+
+
+class AnswerQueryOutput(TypedDict):
+    decomp_answer_results: list[SearchAnswerResults]

backend/danswer/agent_search/core_state.py

@@ -0,0 +1,15 @@
+from typing import TypedDict
+
+from sqlalchemy.orm import Session
+
+from danswer.context.search.models import SearchRequest
+from danswer.llm.interfaces import LLM
+
+
+class PrimaryState(TypedDict, total=False):
+    search_request: SearchRequest
+    primary_llm: LLM
+    fast_llm: LLM
+    # a single session for the entire agent search
+    # is fine if we are only reading
+    db_session: Session

backend/danswer/agent_search/deep_answer/nodes/answer_generation.py

@@ -0,0 +1,114 @@
+from typing import Any
+
+from langchain_core.messages import HumanMessage
+
+from danswer.agent_search.main.states import MainState
+from danswer.agent_search.shared_graph_utils.prompts import COMBINED_CONTEXT
+from danswer.agent_search.shared_graph_utils.prompts import MODIFIED_RAG_PROMPT
+from danswer.agent_search.shared_graph_utils.utils import format_docs
+from danswer.agent_search.shared_graph_utils.utils import normalize_whitespace
+
+
+# aggregate sub questions and answers
+def deep_answer_generation(state: MainState) -> dict[str, Any]:
+    """
+    Generate answer
+
+    Args:
+        state (messages): The current state
+
+    Returns:
+         dict: The updated state with re-phrased question
+    """
+    print("---DEEP GENERATE---")
+
+    question = state["original_question"]
+    docs = state["deduped_retrieval_docs"]
+
+    deep_answer_context = state["core_answer_dynamic_context"]
+
+    print(f"Number of verified retrieval docs - deep: {len(docs)}")
+
+    combined_context = normalize_whitespace(
+        COMBINED_CONTEXT.format(
+            deep_answer_context=deep_answer_context, formated_docs=format_docs(docs)
+        )
+    )
+
+    msg = [
+        HumanMessage(
+            content=MODIFIED_RAG_PROMPT.format(
+                question=question, combined_context=combined_context
+            )
+        )
+    ]
+
+    # Grader
+    model = state["fast_llm"]
+    response = model.invoke(msg)
+
+    return {
+        "deep_answer": response.content,
+    }
+
+
+def final_stuff(state: MainState) -> dict[str, Any]:
+    """
+    Invokes the agent model to generate a response based on the current state. Given
+    the question, it will decide to retrieve using the retriever tool, or simply end.
+
+    Args:
+        state (messages): The current state
+
+    Returns:
+        dict: The updated state with the agent response appended to messages
+    """
+    print("---FINAL---")
+
+    messages = state["log_messages"]
+    time_ordered_messages = [x.pretty_repr() for x in messages]
+    time_ordered_messages.sort()
+
+    print("Message Log:")
+    print("\n".join(time_ordered_messages))
+
+    initial_sub_qas = state["initial_sub_qas"]
+    initial_sub_qa_list = []
+    for initial_sub_qa in initial_sub_qas:
+        if initial_sub_qa["sub_answer_check"] == "yes":
+            initial_sub_qa_list.append(
+                f'  Question:\n  {initial_sub_qa["sub_question"]}\n  --\n  Answer:\n  {initial_sub_qa["sub_answer"]}\n  -----'
+            )
+
+    initial_sub_qa_context = "\n".join(initial_sub_qa_list)
+
+    base_answer = state["base_answer"]
+
+    print(f"Final Base Answer:\n{base_answer}")
+    print("--------------------------------")
+    print(f"Initial Answered Sub Questions:\n{initial_sub_qa_context}")
+    print("--------------------------------")
+
+    if not state.get("deep_answer"):
+        print("No Deep Answer was required")
+        return {}
+
+    deep_answer = state["deep_answer"]
+    sub_qas = state["sub_qas"]
+    sub_qa_list = []
+    for sub_qa in sub_qas:
+        if sub_qa["sub_answer_check"] == "yes":
+            sub_qa_list.append(
+                f'  Question:\n  {sub_qa["sub_question"]}\n  --\n  Answer:\n  {sub_qa["sub_answer"]}\n  -----'
+            )
+
+    sub_qa_context = "\n".join(sub_qa_list)
+
+    print(f"Final Base Answer:\n{base_answer}")
+    print("--------------------------------")
+    print(f"Final Deep Answer:\n{deep_answer}")
+    print("--------------------------------")
+    print("Sub Questions and Answers:")
+    print(sub_qa_context)
+
+    return {}

backend/danswer/agent_search/deep_answer/nodes/deep_decomp.py

@@ -0,0 +1,78 @@
+import json
+import re
+from datetime import datetime
+from typing import Any
+
+from langchain_core.messages import HumanMessage
+
+from danswer.agent_search.main.states import MainState
+from danswer.agent_search.shared_graph_utils.prompts import DEEP_DECOMPOSE_PROMPT
+from danswer.agent_search.shared_graph_utils.utils import format_entity_term_extraction
+from danswer.agent_search.shared_graph_utils.utils import generate_log_message
+
+
+def decompose(state: MainState) -> dict[str, Any]:
+    """ """
+
+    node_start_time = datetime.now()
+
+    question = state["original_question"]
+    base_answer = state["base_answer"]
+
+    # get the entity term extraction dict and properly format it
+    entity_term_extraction_dict = state["retrieved_entities_relationships"][
+        "retrieved_entities_relationships"
+    ]
+
+    entity_term_extraction_str = format_entity_term_extraction(
+        entity_term_extraction_dict
+    )
+
+    initial_question_answers = state["initial_sub_qas"]
+
+    addressed_question_list = [
+        x["sub_question"]
+        for x in initial_question_answers
+        if x["sub_answer_check"] == "yes"
+    ]
+    failed_question_list = [
+        x["sub_question"]
+        for x in initial_question_answers
+        if x["sub_answer_check"] == "no"
+    ]
+
+    msg = [
+        HumanMessage(
+            content=DEEP_DECOMPOSE_PROMPT.format(
+                question=question,
+                entity_term_extraction_str=entity_term_extraction_str,
+                base_answer=base_answer,
+                answered_sub_questions="\n - ".join(addressed_question_list),
+                failed_sub_questions="\n - ".join(failed_question_list),
+            ),
+        )
+    ]
+
+    # Grader
+    model = state["fast_llm"]
+    response = model.invoke(msg)
+
+    cleaned_response = re.sub(r"```json\n|\n```", "", response.pretty_repr())
+    parsed_response = json.loads(cleaned_response)
+
+    sub_questions_dict = {}
+    for sub_question_nr, sub_question_dict in enumerate(
+        parsed_response["sub_questions"]
+    ):
+        sub_question_dict["answered"] = False
+        sub_question_dict["verified"] = False
+        sub_questions_dict[sub_question_nr] = sub_question_dict
+
+    return {
+        "decomposed_sub_questions_dict": sub_questions_dict,
+        "log_messages": generate_log_message(
+            message="deep - decompose",
+            node_start_time=node_start_time,
+            graph_start_time=state["graph_start_time"],
+        ),
+    }

backend/danswer/agent_search/deep_answer/nodes/entity_term_extraction.py

@@ -0,0 +1,40 @@
+import json
+import re
+from typing import Any
+
+from langchain_core.messages import HumanMessage
+from langchain_core.messages import merge_message_runs
+
+from danswer.agent_search.main.states import MainState
+from danswer.agent_search.shared_graph_utils.prompts import ENTITY_TERM_PROMPT
+from danswer.agent_search.shared_graph_utils.utils import format_docs
+
+
+def entity_term_extraction(state: MainState) -> dict[str, Any]:
+    """Extract entities and terms from the question and context"""
+
+    question = state["original_question"]
+    docs = state["deduped_retrieval_docs"]
+
+    doc_context = format_docs(docs)
+
+    msg = [
+        HumanMessage(
+            content=ENTITY_TERM_PROMPT.format(question=question, context=doc_context),
+        )
+    ]
+    fast_llm = state["fast_llm"]
+    # Grader
+    llm_response_list = list(
+        fast_llm.stream(
+            prompt=msg,
+        )
+    )
+    llm_response = merge_message_runs(llm_response_list, chunk_separator="")[0].content
+
+    cleaned_response = re.sub(r"```json\n|\n```", "", llm_response)
+    parsed_response = json.loads(cleaned_response)
+
+    return {
+        "retrieved_entities_relationships": parsed_response,
+    }

backend/danswer/agent_search/deep_answer/nodes/sub_qa_level_aggregator.py

@@ -0,0 +1,30 @@
+from typing import Any
+
+from danswer.agent_search.main.states import MainState
+
+
+# aggregate sub questions and answers
+def sub_qa_level_aggregator(state: MainState) -> dict[str, Any]:
+    sub_qas = state["sub_qas"]
+
+    dynamic_context_list = [
+        "Below you will find useful information to answer the original question:"
+    ]
+    checked_sub_qas = []
+
+    for core_answer_sub_qa in sub_qas:
+        question = core_answer_sub_qa["sub_question"]
+        answer = core_answer_sub_qa["sub_answer"]
+        verified = core_answer_sub_qa["sub_answer_check"]
+
+        if verified == "yes":
+            dynamic_context_list.append(
+                f"Question:\n{question}\n\nAnswer:\n{answer}\n\n---\n\n"
+            )
+            checked_sub_qas.append({"sub_question": question, "sub_answer": answer})
+    dynamic_context = "\n".join(dynamic_context_list)
+
+    return {
+        "core_answer_dynamic_context": dynamic_context,
+        "checked_sub_qas": checked_sub_qas,
+    }

backend/danswer/agent_search/deep_answer/nodes/sub_qa_manager.py

@@ -0,0 +1,19 @@
+from typing import Any
+
+from danswer.agent_search.main.states import MainState
+
+
+def sub_qa_manager(state: MainState) -> dict[str, Any]:
+    """ """
+
+    sub_questions_dict = state["decomposed_sub_questions_dict"]
+
+    sub_questions = {}
+
+    for sub_question_nr, sub_question_dict in sub_questions_dict.items():
+        sub_questions[sub_question_nr] = sub_question_dict["sub_question"]
+
+    return {
+        "sub_questions": sub_questions,
+        "num_new_question_iterations": 0,
+    }

backend/danswer/agent_search/expanded_retrieval/edges.py

@@ -0,0 +1,44 @@
+from collections.abc import Hashable
+
+from langchain_core.messages import HumanMessage
+from langchain_core.messages import merge_message_runs
+from langgraph.types import Send
+
+from danswer.agent_search.expanded_retrieval.nodes.doc_retrieval import RetrieveInput
+from danswer.agent_search.expanded_retrieval.states import ExpandedRetrievalInput
+from danswer.agent_search.shared_graph_utils.prompts import REWRITE_PROMPT_MULTI
+from danswer.llm.interfaces import LLM
+
+
+def parallel_retrieval_edge(state: ExpandedRetrievalInput) -> list[Send | Hashable]:
+    print(f"parallel_retrieval_edge state: {state.keys()}")
+
+    # This should be better...
+    question = state.get("query_to_answer") or state["search_request"].query
+    llm: LLM = state["fast_llm"]
+
+    msg = [
+        HumanMessage(
+            content=REWRITE_PROMPT_MULTI.format(question=question),
+        )
+    ]
+    llm_response_list = list(
+        llm.stream(
+            prompt=msg,
+        )
+    )
+    llm_response = merge_message_runs(llm_response_list, chunk_separator="")[0].content
+
+    print(f"llm_response: {llm_response}")
+
+    rewritten_queries = llm_response.split("\n")
+
+    print(f"rewritten_queries: {rewritten_queries}")
+
+    return [
+        Send(
+            "doc_retrieval",
+            RetrieveInput(query_to_retrieve=query, **state),
+        )
+        for query in rewritten_queries
+    ]

backend/danswer/agent_search/expanded_retrieval/graph_builder.py

@@ -0,0 +1,88 @@
+from langgraph.graph import END
+from langgraph.graph import START
+from langgraph.graph import StateGraph
+
+from danswer.agent_search.expanded_retrieval.edges import parallel_retrieval_edge
+from danswer.agent_search.expanded_retrieval.nodes.doc_reranking import doc_reranking
+from danswer.agent_search.expanded_retrieval.nodes.doc_retrieval import doc_retrieval
+from danswer.agent_search.expanded_retrieval.nodes.doc_verification import (
+    doc_verification,
+)
+from danswer.agent_search.expanded_retrieval.nodes.verification_kickoff import (
+    verification_kickoff,
+)
+from danswer.agent_search.expanded_retrieval.states import ExpandedRetrievalInput
+from danswer.agent_search.expanded_retrieval.states import ExpandedRetrievalOutput
+from danswer.agent_search.expanded_retrieval.states import ExpandedRetrievalState
+
+
+def expanded_retrieval_graph_builder() -> StateGraph:
+    graph = StateGraph(
+        state_schema=ExpandedRetrievalState,
+        input=ExpandedRetrievalInput,
+        output=ExpandedRetrievalOutput,
+    )
+
+    ### Add nodes ###
+
+    graph.add_node(
+        node="doc_retrieval",
+        action=doc_retrieval,
+    )
+    graph.add_node(
+        node="verification_kickoff",
+        action=verification_kickoff,
+    )
+    graph.add_node(
+        node="doc_verification",
+        action=doc_verification,
+    )
+    graph.add_node(
+        node="doc_reranking",
+        action=doc_reranking,
+    )
+
+    ### Add edges ###
+
+    graph.add_conditional_edges(
+        source=START,
+        path=parallel_retrieval_edge,
+        path_map=["doc_retrieval"],
+    )
+    graph.add_edge(
+        start_key="doc_retrieval",
+        end_key="verification_kickoff",
+    )
+    graph.add_edge(
+        start_key="doc_verification",
+        end_key="doc_reranking",
+    )
+    graph.add_edge(
+        start_key="doc_reranking",
+        end_key=END,
+    )
+
+    return graph
+
+
+if __name__ == "__main__":
+    from danswer.db.engine import get_session_context_manager
+    from danswer.llm.factory import get_default_llms
+    from danswer.context.search.models import SearchRequest
+
+    graph = expanded_retrieval_graph_builder()
+    compiled_graph = graph.compile()
+    primary_llm, fast_llm = get_default_llms()
+    search_request = SearchRequest(
+        query="Who made Excel and what other products did they make?",
+    )
+    with get_session_context_manager() as db_session:
+        inputs = ExpandedRetrievalInput(
+            search_request=search_request,
+            primary_llm=primary_llm,
+            fast_llm=fast_llm,
+            db_session=db_session,
+            query_to_answer="Who made Excel?",
+        )
+        for thing in compiled_graph.stream(inputs, debug=True):
+            print(thing)

backend/danswer/agent_search/expanded_retrieval/nodes/doc_reranking.py

@@ -0,0 +1,11 @@
+from danswer.agent_search.expanded_retrieval.states import DocRerankingOutput
+from danswer.agent_search.expanded_retrieval.states import ExpandedRetrievalState
+
+
+def doc_reranking(state: ExpandedRetrievalState) -> DocRerankingOutput:
+    print(f"doc_reranking state: {state.keys()}")
+
+    verified_documents = state["verified_documents"]
+    reranked_documents = verified_documents
+
+    return DocRerankingOutput(reranked_documents=reranked_documents)

backend/danswer/agent_search/expanded_retrieval/nodes/doc_retrieval.py

@@ -0,0 +1,47 @@
+from danswer.agent_search.expanded_retrieval.states import DocRetrievalOutput
+from danswer.agent_search.expanded_retrieval.states import ExpandedRetrievalState
+from danswer.context.search.models import InferenceSection
+from danswer.context.search.models import SearchRequest
+from danswer.context.search.pipeline import SearchPipeline
+from danswer.db.engine import get_session_context_manager
+
+
+class RetrieveInput(ExpandedRetrievalState):
+    query_to_retrieve: str
+
+
+def doc_retrieval(state: RetrieveInput) -> DocRetrievalOutput:
+    # def doc_retrieval(state: RetrieveInput) -> Command[Literal["doc_verification"]]:
+    """
+    Retrieve documents
+
+    Args:
+        state (dict): The current graph state
+
+    Returns:
+        state (dict): New key added to state, documents, that contains retrieved documents
+    """
+    print(f"doc_retrieval state: {state.keys()}")
+
+    state["query_to_retrieve"]
+
+    documents: list[InferenceSection] = []
+    llm = state["primary_llm"]
+    fast_llm = state["fast_llm"]
+    # db_session = state["db_session"]
+    query_to_retrieve = state["search_request"].query
+    with get_session_context_manager() as db_session1:
+        documents = SearchPipeline(
+            search_request=SearchRequest(
+                query=query_to_retrieve,
+            ),
+            user=None,
+            llm=llm,
+            fast_llm=fast_llm,
+            db_session=db_session1,
+        ).reranked_sections
+
+    print(f"retrieved documents: {len(documents)}")
+    return DocRetrievalOutput(
+        retrieved_documents=documents,
+    )

backend/danswer/agent_search/expanded_retrieval/nodes/doc_verification.py

@@ -0,0 +1,60 @@
+from langchain_core.messages import HumanMessage
+from langchain_core.messages import merge_message_runs
+
+from danswer.agent_search.expanded_retrieval.states import DocVerificationOutput
+from danswer.agent_search.expanded_retrieval.states import ExpandedRetrievalState
+from danswer.agent_search.shared_graph_utils.models import BinaryDecision
+from danswer.agent_search.shared_graph_utils.prompts import VERIFIER_PROMPT
+from danswer.context.search.models import InferenceSection
+
+
+class DocVerificationInput(ExpandedRetrievalState, total=True):
+    doc_to_verify: InferenceSection
+
+
+def doc_verification(state: DocVerificationInput) -> DocVerificationOutput:
+    """
+    Check whether the document is relevant for the original user question
+
+    Args:
+        state (VerifierState): The current state
+
+    Returns:
+        dict: ict: The updated state with the final decision
+    """
+
+    print(f"doc_verification state: {state.keys()}")
+
+    original_query = state["search_request"].query
+    doc_to_verify = state["doc_to_verify"]
+    document_content = doc_to_verify.combined_content
+
+    msg = [
+        HumanMessage(
+            content=VERIFIER_PROMPT.format(
+                question=original_query, document_content=document_content
+            )
+        )
+    ]
+
+    fast_llm = state["fast_llm"]
+    response = list(
+        fast_llm.stream(
+            prompt=msg,
+        )
+    )
+
+    response_string = merge_message_runs(response, chunk_separator="")[0].content
+    # Convert string response to proper dictionary format
+    decision_dict = {"decision": response_string.lower()}
+    formatted_response = BinaryDecision.model_validate(decision_dict)
+
+    print(f"Verdict: {formatted_response.decision}")
+
+    verified_documents = []
+    if formatted_response.decision == "yes":
+        verified_documents.append(doc_to_verify)
+
+    return DocVerificationOutput(
+        verified_documents=verified_documents,
+    )

backend/danswer/agent_search/expanded_retrieval/nodes/verification_kickoff.py

@@ -0,0 +1,27 @@
+from typing import Literal
+
+from langgraph.types import Command
+from langgraph.types import Send
+
+from danswer.agent_search.expanded_retrieval.nodes.doc_verification import (
+    DocVerificationInput,
+)
+from danswer.agent_search.expanded_retrieval.states import ExpandedRetrievalState
+
+
+def verification_kickoff(
+    state: ExpandedRetrievalState,
+) -> Command[Literal["doc_verification"]]:
+    print(f"verification_kickoff state: {state.keys()}")
+
+    documents = state["retrieved_documents"]
+    return Command(
+        update={},
+        goto=[
+            Send(
+                node="doc_verification",
+                arg=DocVerificationInput(doc_to_verify=doc, **state),
+            )
+            for doc in documents
+        ],
+    )

backend/danswer/agent_search/expanded_retrieval/states.py

@@ -0,0 +1,36 @@
+from typing import Annotated
+from typing import TypedDict
+
+from danswer.agent_search.core_state import PrimaryState
+from danswer.agent_search.shared_graph_utils.operators import dedup_inference_sections
+from danswer.context.search.models import InferenceSection
+
+
+class DocRetrievalOutput(TypedDict, total=False):
+    retrieved_documents: Annotated[list[InferenceSection], dedup_inference_sections]
+
+
+class DocVerificationOutput(TypedDict, total=False):
+    verified_documents: Annotated[list[InferenceSection], dedup_inference_sections]
+
+
+class DocRerankingOutput(TypedDict, total=False):
+    reranked_documents: Annotated[list[InferenceSection], dedup_inference_sections]
+
+
+class ExpandedRetrievalState(
+    PrimaryState,
+    DocRetrievalOutput,
+    DocVerificationOutput,
+    DocRerankingOutput,
+    total=True,
+):
+    query_to_answer: str
+
+
+class ExpandedRetrievalInput(PrimaryState, total=True):
+    query_to_answer: str
+
+
+class ExpandedRetrievalOutput(TypedDict):
+    reordered_documents: Annotated[list[InferenceSection], dedup_inference_sections]

backend/danswer/agent_search/main/edges.py

@@ -0,0 +1,61 @@
+from collections.abc import Hashable
+
+from langgraph.types import Send
+
+from danswer.agent_search.answer_query.states import AnswerQueryInput
+from danswer.agent_search.main.states import MainState
+
+
+def parallelize_decompozed_answer_queries(state: MainState) -> list[Send | Hashable]:
+    return [
+        Send(
+            "answer_query",
+            AnswerQueryInput(
+                **state,
+                query_to_answer=query,
+            ),
+        )
+        for query in state["initial_decomp_queries"]
+    ]
+
+
+# def continue_to_answer_sub_questions(state: QAState) -> Union[Hashable, list[Hashable]]:
+#     # Routes re-written queries to the (parallel) retrieval steps
+#     # Notice the 'Send()' API that takes care of the parallelization
+#     return [
+#         Send(
+#             "sub_answers_graph",
+#             ResearchQAState(
+#                 sub_question=sub_question["sub_question_str"],
+#                 sub_question_nr=sub_question["sub_question_nr"],
+#                 graph_start_time=state["graph_start_time"],
+#                 primary_llm=state["primary_llm"],
+#                 fast_llm=state["fast_llm"],
+#             ),
+#         )
+#         for sub_question in state["sub_questions"]
+#     ]
+
+
+# def continue_to_deep_answer(state: QAState) -> Union[Hashable, list[Hashable]]:
+#     print("---GO TO DEEP ANSWER OR END---")
+
+#     base_answer = state["base_answer"]
+
+#     question = state["original_question"]
+
+#     BASE_CHECK_MESSAGE = [
+#         HumanMessage(
+#             content=BASE_CHECK_PROMPT.format(question=question, base_answer=base_answer)
+#         )
+#     ]
+
+#     model = state["fast_llm"]
+#     response = model.invoke(BASE_CHECK_MESSAGE)
+
+#     print(f"CAN WE CONTINUE W/O GENERATING A DEEP ANSWER? - {response.pretty_repr()}")
+
+#     if response.pretty_repr() == "no":
+#         return "decompose"
+#     else:
+#         return "end"

backend/danswer/agent_search/main/graph_builder.py

@@ -0,0 +1,98 @@
+from langgraph.graph import END
+from langgraph.graph import START
+from langgraph.graph import StateGraph
+
+from danswer.agent_search.answer_query.graph_builder import answer_query_graph_builder
+from danswer.agent_search.expanded_retrieval.graph_builder import (
+    expanded_retrieval_graph_builder,
+)
+from danswer.agent_search.main.edges import parallelize_decompozed_answer_queries
+from danswer.agent_search.main.nodes.base_decomp import main_decomp_base
+from danswer.agent_search.main.nodes.generate_initial_answer import (
+    generate_initial_answer,
+)
+from danswer.agent_search.main.states import MainInput
+from danswer.agent_search.main.states import MainState
+
+
+def main_graph_builder() -> StateGraph:
+    graph = StateGraph(
+        state_schema=MainState,
+        input=MainInput,
+    )
+
+    ### Add nodes ###
+
+    graph.add_node(
+        node="base_decomp",
+        action=main_decomp_base,
+    )
+    answer_query_subgraph = answer_query_graph_builder().compile()
+    graph.add_node(
+        node="answer_query",
+        action=answer_query_subgraph,
+    )
+    expanded_retrieval_subgraph = expanded_retrieval_graph_builder().compile()
+    graph.add_node(
+        node="expanded_retrieval",
+        action=expanded_retrieval_subgraph,
+    )
+    graph.add_node(
+        node="generate_initial_answer",
+        action=generate_initial_answer,
+    )
+
+    ### Add edges ###
+    graph.add_edge(
+        start_key=START,
+        end_key="expanded_retrieval",
+    )
+
+    graph.add_edge(
+        start_key=START,
+        end_key="base_decomp",
+    )
+    graph.add_conditional_edges(
+        source="base_decomp",
+        path=parallelize_decompozed_answer_queries,
+        path_map=["answer_query"],
+    )
+    graph.add_edge(
+        start_key=["answer_query", "expanded_retrieval"],
+        end_key="generate_initial_answer",
+    )
+    graph.add_edge(
+        start_key="generate_initial_answer",
+        end_key=END,
+    )
+
+    return graph
+
+
+if __name__ == "__main__":
+    from danswer.db.engine import get_session_context_manager
+    from danswer.llm.factory import get_default_llms
+    from danswer.context.search.models import SearchRequest
+
+    graph = main_graph_builder()
+    compiled_graph = graph.compile()
+    primary_llm, fast_llm = get_default_llms()
+    search_request = SearchRequest(
+        query="If i am familiar with the function that I need, how can I type it into a cell?",
+    )
+    with get_session_context_manager() as db_session:
+        inputs = MainInput(
+            search_request=search_request,
+            primary_llm=primary_llm,
+            fast_llm=fast_llm,
+            db_session=db_session,
+        )
+        for thing in compiled_graph.stream(
+            input=inputs,
+            # stream_mode="debug",
+            # debug=True,
+            subgraphs=True,
+        ):
+            # print(thing)
+            print()
+            print()

backend/danswer/agent_search/main/nodes/base_decomp.py

@@ -0,0 +1,31 @@
+from langchain_core.messages import HumanMessage
+
+from danswer.agent_search.main.states import BaseDecompOutput
+from danswer.agent_search.main.states import MainState
+from danswer.agent_search.shared_graph_utils.prompts import INITIAL_DECOMPOSITION_PROMPT
+from danswer.agent_search.shared_graph_utils.utils import clean_and_parse_list_string
+
+
+def main_decomp_base(state: MainState) -> BaseDecompOutput:
+    question = state["search_request"].query
+
+    msg = [
+        HumanMessage(
+            content=INITIAL_DECOMPOSITION_PROMPT.format(question=question),
+        )
+    ]
+
+    # Get the rewritten queries in a defined format
+    model = state["fast_llm"]
+    response = model.invoke(msg)
+
+    content = response.pretty_repr()
+    list_of_subquestions = clean_and_parse_list_string(content)
+
+    decomp_list: list[str] = [
+        sub_question["sub_question"].strip() for sub_question in list_of_subquestions
+    ]
+
+    return BaseDecompOutput(
+        initial_decomp_queries=decomp_list,
+    )

backend/danswer/agent_search/main/nodes/generate_initial_answer.py

@@ -0,0 +1,53 @@
+from langchain_core.messages import HumanMessage
+
+from danswer.agent_search.main.states import InitialAnswerOutput
+from danswer.agent_search.main.states import MainState
+from danswer.agent_search.shared_graph_utils.prompts import INITIAL_RAG_PROMPT
+from danswer.agent_search.shared_graph_utils.utils import format_docs
+
+
+def generate_initial_answer(state: MainState) -> InitialAnswerOutput:
+    print("---GENERATE INITIAL---")
+
+    question = state["search_request"].query
+    docs = state["documents"]
+
+    decomp_answer_results = state["decomp_answer_results"]
+
+    good_qa_list: list[str] = []
+
+    _SUB_QUESTION_ANSWER_TEMPLATE = """
+    Sub-Question:\n  - {sub_question}\n  --\nAnswer:\n  - {sub_answer}\n\n
+    """
+    for decomp_answer_result in decomp_answer_results:
+        if (
+            decomp_answer_result.quality.lower() == "yes"
+            and len(decomp_answer_result.answer) > 0
+            and decomp_answer_result.answer != "I don't know"
+        ):
+            good_qa_list.append(
+                _SUB_QUESTION_ANSWER_TEMPLATE.format(
+                    sub_question=decomp_answer_result.query,
+                    sub_answer=decomp_answer_result.answer,
+                )
+            )
+
+    sub_question_answer_str = "\n\n------\n\n".join(good_qa_list)
+
+    msg = [
+        HumanMessage(
+            content=INITIAL_RAG_PROMPT.format(
+                question=question,
+                context=format_docs(docs),
+                answered_sub_questions=sub_question_answer_str,
+            )
+        )
+    ]
+
+    # Grader
+    model = state["fast_llm"]
+    response = model.invoke(msg)
+    answer = response.pretty_repr()
+
+    print(answer)
+    return InitialAnswerOutput(initial_answer=answer)

backend/danswer/agent_search/main/states.py

@@ -0,0 +1,37 @@
+from operator import add
+from typing import Annotated
+from typing import TypedDict
+
+from danswer.agent_search.answer_query.states import SearchAnswerResults
+from danswer.agent_search.core_state import PrimaryState
+from danswer.agent_search.shared_graph_utils.operators import dedup_inference_sections
+from danswer.context.search.models import InferenceSection
+
+
+class BaseDecompOutput(TypedDict, total=False):
+    initial_decomp_queries: list[str]
+
+
+class InitialAnswerOutput(TypedDict, total=False):
+    initial_answer: str
+
+
+class MainState(
+    PrimaryState,
+    BaseDecompOutput,
+    InitialAnswerOutput,
+    total=True,
+):
+    documents: Annotated[list[InferenceSection], dedup_inference_sections]
+    decomp_answer_results: Annotated[list[SearchAnswerResults], add]
+
+
+class MainInput(PrimaryState, total=True):
+    pass
+
+
+class MainOutput(TypedDict):
+    """
+    This is not used because defining the output only matters for filtering the output of
+      a .invoke() call but we are streaming so we just yield the entire state.
+    """

backend/danswer/agent_search/run_graph.py

@@ -0,0 +1,27 @@
+from danswer.agent_search.primary_graph.graph_builder import build_core_graph
+from danswer.llm.answering.answer import AnswerStream
+from danswer.llm.interfaces import LLM
+from danswer.tools.tool import Tool
+
+
+def run_graph(
+    query: str,
+    llm: LLM,
+    tools: list[Tool],
+) -> AnswerStream:
+    graph = build_core_graph()
+
+    inputs = {
+        "original_query": query,
+        "messages": [],
+        "tools": tools,
+        "llm": llm,
+    }
+    compiled_graph = graph.compile()
+    output = compiled_graph.invoke(input=inputs)
+    yield from output
+
+
+if __name__ == "__main__":
+    pass
+    # run_graph("What is the capital of France?", llm, [])

backend/danswer/agent_search/shared_graph_utils/models.py

@@ -0,0 +1,12 @@
+from typing import Literal
+
+from pydantic import BaseModel
+
+
+# Pydantic models for structured outputs
+class RewrittenQueries(BaseModel):
+    rewritten_queries: list[str]
+
+
+class BinaryDecision(BaseModel):
+    decision: Literal["yes", "no"]

backend/danswer/agent_search/shared_graph_utils/operators.py

@@ -0,0 +1,9 @@
+from danswer.context.search.models import InferenceSection
+from danswer.llm.answering.prune_and_merge import _merge_sections
+
+
+def dedup_inference_sections(
+    list1: list[InferenceSection], list2: list[InferenceSection]
+) -> list[InferenceSection]:
+    deduped = _merge_sections(list1 + list2)
+    return deduped

backend/danswer/agent_search/shared_graph_utils/prompts.py

@@ -0,0 +1,427 @@
+REWRITE_PROMPT_MULTI_ORIGINAL = """ \n
+    Please convert an initial user question into a 2-3 more appropriate short and pointed search queries for retrievel from a
+    document store. Particularly, try to think about resolving ambiguities and make the search queries more specific,
+    enabling the system to search more broadly.
+    Also, try to make the search queries not redundant, i.e. not too similar! \n\n
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+    Formulate the queries separated by '--' (Do not say 'Query 1: ...', just write the querytext): """
+
+REWRITE_PROMPT_MULTI = """ \n
+    Please create a list of 2-3 sample documents that could answer an original question. Each document
+    should be about as long as the original question. \n
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+    Formulate the sample documents separated by '--' (Do not say 'Document 1: ...', just write the text): """
+
+BASE_RAG_PROMPT = """ \n
+    You are an assistant for question-answering tasks. Use the context provided below - and only the
+    provided context - to answer the question. If you don't know the answer or if the provided context is
+    empty, just say "I don't know". Do not use your internal knowledge!
+
+    Again, only use the provided context and do not use your internal knowledge! If you cannot answer the
+    question based on the context, say "I don't know". It is a matter of life and death that you do NOT
+    use your internal knowledge, just the provided information!
+
+    Use three sentences maximum and keep the answer concise.
+    answer concise.\nQuestion:\n {question} \nContext:\n {context} \n\n
+    \n\n
+    Answer:"""
+
+BASE_CHECK_PROMPT = """ \n
+    Please check whether 1) the suggested answer seems to fully address the original question AND 2)the
+    original question requests a simple, factual answer, and there are no ambiguities, judgements,
+    aggregations, or any other complications that may require extra context. (I.e., if the question is
+    somewhat addressed, but the answer would benefit from more context, then answer with 'no'.)
+
+    Please only answer with 'yes' or 'no' \n
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+    Here is the proposed answer:
+    \n ------- \n
+    {base_answer}
+    \n ------- \n
+    Please answer with yes or no:"""
+
+VERIFIER_PROMPT = """ \n
+    Please check whether the document seems to be relevant for the answer of the question. Please
+    only answer with 'yes' or 'no' \n
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+    Here is the document text:
+    \n ------- \n
+    {document_content}
+    \n ------- \n
+    Please answer with yes or no:"""
+
+INITIAL_DECOMPOSITION_PROMPT_BASIC = """ \n
+    Please decompose an initial user question into not more than 4 appropriate sub-questions that help to
+    answer the original question. The purpose for this decomposition is to isolate individulal entities
+    (i.e., 'compare sales of company A and company B' -> 'what are sales for company A' + 'what are sales
+    for company B'), split ambiguous terms (i.e., 'what is our success with company A' -> 'what are our
+    sales with company A' + 'what is our market share with company A' + 'is company A a reference customer
+    for us'), etc. Each sub-question should be realistically be answerable by a good RAG system. \n
+
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+
+    Please formulate your answer as a list of subquestions:
+
+    Answer:
+    """
+
+REWRITE_PROMPT_SINGLE = """ \n
+    Please convert an initial user question into a more appropriate search query for retrievel from a
+    document store. \n
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+
+    Formulate the query: """
+
+MODIFIED_RAG_PROMPT = """You are an assistant for question-answering tasks. Use the context provided below
+    - and only this context - to answer the question. If you don't know the answer, just say "I don't know".
+    Use three sentences maximum and keep the answer concise.
+    Pay also particular attention to the sub-questions and their answers, at least it may enrich the answer.
+    Again, only use the provided context and do not use your internal knowledge! If you cannot answer the
+    question based on the context, say "I don't know". It is a matter of life and death that you do NOT
+    use your internal knowledge, just the provided information!
+
+    \nQuestion: {question}
+    \nContext: {combined_context} \n
+
+    Answer:"""
+
+ORIG_DEEP_DECOMPOSE_PROMPT = """ \n
+    An initial user question needs to be answered. An initial answer has been provided but it wasn't quite
+    good enough. Also, some sub-questions had been answered and this information has been used to provide
+    the initial answer. Some other subquestions may have been suggested based on little knowledge, but they
+    were not directly answerable. Also, some entities, relationships and terms are givenm to you so that
+    you have an idea of how the avaiolable data looks like.
+
+    Your role is to generate 3-5 new sub-questions that would help to answer the initial question,
+    considering:
+
+    1) The initial question
+    2) The initial answer that was found to be unsatisfactory
+    3) The sub-questions that were answered
+    4) The sub-questions that were suggested but not answered
+    5) The entities, relationships and terms that were extracted from the context
+
+    The individual questions should be answerable by a good RAG system.
+    So a good idea would be to use the sub-questions to resolve ambiguities and/or to separate the
+    question for different entities that may be involved in the original question, but in a way that does
+    not duplicate questions that were already tried.
+
+    Additional Guidelines:
+    - The sub-questions should be specific to the question and provide richer context for the question,
+    resolve ambiguities, or address shortcoming of the initial answer
+    - Each sub-question - when answered - should be relevant for the answer to the original question
+    - The sub-questions should be free from comparisions, ambiguities,judgements, aggregations, or any
+    other complications that may require extra context.
+    - The sub-questions MUST have the full context of the original question so that it can be executed by
+    a RAG system independently without the original question available
+      (Example:
+        - initial question: "What is the capital of France?"
+        - bad sub-question: "What is the name of the river there?"
+        - good sub-question: "What is the name of the river that flows through Paris?"
+    - For each sub-question, please provide a short explanation for why it is a good sub-question. So
+    generate a list of dictionaries with the following format:
+      [{{"sub_question": <sub-question>, "explanation": <explanation>, "search_term": <rewrite the
+      sub-question using as a search phrase for the document store>}}, ...]
+
+    \n\n
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+
+    Here is the initial sub-optimal answer:
+    \n ------- \n
+    {base_answer}
+    \n ------- \n
+
+    Here are the sub-questions that were answered:
+    \n ------- \n
+    {answered_sub_questions}
+    \n ------- \n
+
+    Here are the sub-questions that were suggested but not answered:
+    \n ------- \n
+    {failed_sub_questions}
+    \n ------- \n
+
+    And here are the entities, relationships and terms extracted from the context:
+    \n ------- \n
+    {entity_term_extraction_str}
+    \n ------- \n
+
+   Please generate the list of good, fully contextualized sub-questions that would help to address the
+   main question. Again, please find questions that are NOT overlapping too much with the already answered
+   sub-questions or those that already were suggested and failed.
+   In other words - what can we try in addition to what has been tried so far?
+
+   Please think through it step by step and then generate the list of json dictionaries with the following
+   format:
+
+   {{"sub_questions": [{{"sub_question": <sub-question>,
+        "explanation": <explanation>,
+        "search_term": <rewrite the sub-question using as a search phrase for the document store>}},
+        ...]}} """
+
+DEEP_DECOMPOSE_PROMPT = """ \n
+    An initial user question needs to be answered. An initial answer has been provided but it wasn't quite
+    good enough. Also, some sub-questions had been answered and this information has been used to provide
+    the initial answer. Some other subquestions may have been suggested based on little knowledge, but they
+    were not directly answerable. Also, some entities, relationships and terms are givenm to you so that
+    you have an idea of how the avaiolable data looks like.
+
+    Your role is to generate 4-6 new sub-questions that would help to answer the initial question,
+    considering:
+
+    1) The initial question
+    2) The initial answer that was found to be unsatisfactory
+    3) The sub-questions that were answered
+    4) The sub-questions that were suggested but not answered
+    5) The entities, relationships and terms that were extracted from the context
+
+    The individual questions should be answerable by a good RAG system.
+    So a good idea would be to use the sub-questions to resolve ambiguities and/or to separate the
+    question for different entities that may be involved in the original question, but in a way that does
+    not duplicate questions that were already tried.
+
+    Additional Guidelines:
+    - The sub-questions should be specific to the question and provide richer context for the question,
+    resolve ambiguities, or address shortcoming of the initial answer
+    - Each sub-question - when answered - should be relevant for the answer to the original question
+    - The sub-questions should be free from comparisions, ambiguities,judgements, aggregations, or any
+    other complications that may require extra context.
+    - The sub-questions MUST have the full context of the original question so that it can be executed by
+    a RAG system independently without the original question available
+      (Example:
+        - initial question: "What is the capital of France?"
+        - bad sub-question: "What is the name of the river there?"
+        - good sub-question: "What is the name of the river that flows through Paris?"
+    - For each sub-question, please also provide a search term that can be used to retrieve relevant
+    documents from a document store.
+    \n\n
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+
+    Here is the initial sub-optimal answer:
+    \n ------- \n
+    {base_answer}
+    \n ------- \n
+
+    Here are the sub-questions that were answered:
+    \n ------- \n
+    {answered_sub_questions}
+    \n ------- \n
+
+    Here are the sub-questions that were suggested but not answered:
+    \n ------- \n
+    {failed_sub_questions}
+    \n ------- \n
+
+    And here are the entities, relationships and terms extracted from the context:
+    \n ------- \n
+    {entity_term_extraction_str}
+    \n ------- \n
+
+   Please generate the list of good, fully contextualized sub-questions that would help to address the
+   main question. Again, please find questions that are NOT overlapping too much with the already answered
+   sub-questions or those that already were suggested and failed.
+   In other words - what can we try in addition to what has been tried so far?
+
+   Generate the list of json dictionaries with the following format:
+
+   {{"sub_questions": [{{"sub_question": <sub-question>,
+        "search_term": <rewrite the sub-question using as a search phrase for the document store>}},
+        ...]}} """
+
+DECOMPOSE_PROMPT = """ \n
+    For an initial user question, please generate at 5-10 individual sub-questions whose answers would help
+    \n to answer the initial question. The individual questions should be answerable by a good RAG system.
+    So a good idea would be to \n use the sub-questions to resolve ambiguities and/or to separate the
+    question for different entities that may be involved in the original question.
+
+    In order to arrive at meaningful sub-questions, please also consider the context retrieved from the
+    document store, expressed as entities, relationships and terms. You can also think about the types
+    mentioned in brackets
+
+    Guidelines:
+    - The sub-questions should be specific to the question and provide richer context for the question,
+    and or resolve ambiguities
+    - Each sub-question - when answered - should be relevant for the answer to the original question
+    - The sub-questions should be free from comparisions, ambiguities,judgements, aggregations, or any
+    other complications that may require extra context.
+    - The sub-questions MUST have the full context of the original question so that it can be executed by
+    a RAG system independently without the original question available
+      (Example:
+        - initial question: "What is the capital of France?"
+        - bad sub-question: "What is the name of the river there?"
+        - good sub-question: "What is the name of the river that flows through Paris?"
+    - For each sub-question, please provide a short explanation for why it is a good sub-question. So
+    generate a list of dictionaries with the following format:
+      [{{"sub_question": <sub-question>, "explanation": <explanation>, "search_term": <rewrite the
+      sub-question using as a search phrase for the document store>}}, ...]
+
+    \n\n
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+
+    And here are the entities, relationships and terms extracted from the context:
+    \n ------- \n
+    {entity_term_extraction_str}
+    \n ------- \n
+
+   Please generate the list of good, fully contextualized sub-questions that would help to address the
+   main question. Don't be too specific unless the original question is specific.
+   Please think through it step by step and then generate the list of json dictionaries with the following
+   format:
+   {{"sub_questions": [{{"sub_question": <sub-question>,
+        "explanation": <explanation>,
+        "search_term": <rewrite the sub-question using as a search phrase for the document store>}},
+        ...]}} """
+
+#### Consolidations
+COMBINED_CONTEXT = """-------
+    Below you will find useful information to answer the original question. First, you see a number of
+    sub-questions with their answers. This information should be considered to be more focussed and
+    somewhat more specific to the original question as it tries to contextualized facts.
+    After that will see the documents that were considered to be relevant to answer the original question.
+
+    Here are the sub-questions and their answers:
+    \n\n {deep_answer_context} \n\n
+    \n\n Here are the documents that were considered to be relevant to answer the original question:
+    \n\n {formated_docs} \n\n
+    ----------------
+    """
+
+SUB_QUESTION_EXPLANATION_RANKER_PROMPT = """-------
+    Below you will find a question that we ultimately want to answer (the original question) and a list of
+    motivations in arbitrary order for generated sub-questions that are supposed to help us answering the
+    original question. The motivations are formatted as <motivation number>:  <motivation explanation>.
+    (Again, the numbering is arbitrary and does not necessarily mean that 1 is the most relevant
+    motivation and 2 is less relevant.)
+
+    Please rank the motivations in order of relevance for answering the original question. Also, try to
+    ensure that the top questions do not duplicate too much, i.e. that they are not too similar.
+    Ultimately, create a list with the motivation numbers where the number of the most relevant
+    motivations comes first.
+
+    Here is the original question:
+    \n\n {original_question} \n\n
+    \n\n Here is the list of sub-question motivations:
+    \n\n {sub_question_explanations} \n\n
+    ----------------
+
+    Please think step by step and then generate the ranked list of motivations.
+
+    Please format your answer as a json object in the following format:
+    {{"reasonning": <explain your reasoning for the ranking>,
+      "ranked_motivations": <ranked list of motivation numbers>}}
+    """
+
+
+INITIAL_DECOMPOSITION_PROMPT = """ \n
+    Please decompose an initial user question into 2 or 3 appropriate sub-questions that help to
+    answer the original question. The purpose for this decomposition is to isolate individulal entities
+    (i.e., 'compare sales of company A and company B' -> 'what are sales for company A' + 'what are sales
+    for company B'), split ambiguous terms (i.e., 'what is our success with company A' -> 'what are our
+    sales with company A' + 'what is our market share with company A' + 'is company A a reference customer
+    for us'), etc. Each sub-question should be realistically be answerable by a good RAG system. \n
+
+    For each sub-question, please also create one search term that can be used to retrieve relevant
+    documents from a document store.
+
+    Here is the initial question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+
+    Please formulate your answer as a list of json objects with the following format:
+
+   [{{"sub_question": <sub-question>, "search_term": <search term>}}, ...]
+
+    Answer:
+    """
+
+INITIAL_RAG_PROMPT = """ \n
+    You are an assistant for question-answering tasks. Use the information provided below - and only the
+    provided information - to answer the provided question.
+
+    The information provided below consists of:
+     1) a number of answered sub-questions - these are very important(!) and definitely should be
+     considered to answer the question.
+     2) a number of documents that were also deemed relevant for the question.
+
+    If you don't know the answer or if the provided information is empty or insufficient, just say
+    "I don't know". Do not use your internal knowledge!
+
+    Again, only use the provided informationand do not use your internal knowledge! It is a matter of life
+    and death that you do NOT use your internal knowledge, just the provided information!
+
+    Try to keep your answer concise.
+
+    And here is the question and the provided information:
+    \n
+    \nQuestion:\n {question}
+
+    \nAnswered Sub-questions:\n {answered_sub_questions}
+
+    \nContext:\n {context} \n\n
+    \n\n
+
+    Answer:"""
+
+ENTITY_TERM_PROMPT = """ \n
+    Based on the original question and the context retieved from a dataset, please generate a list of
+    entities (e.g. companies, organizations, industries, products, locations, etc.), terms and concepts
+    (e.g. sales, revenue, etc.) that are relevant for the question, plus their relations to each other.
+
+    \n\n
+    Here is the original question:
+    \n ------- \n
+    {question}
+    \n ------- \n
+   And here is the context retrieved:
+    \n ------- \n
+    {context}
+    \n ------- \n
+
+    Please format your answer as a json object in the following format:
+
+    {{"retrieved_entities_relationships": {{
+        "entities": [{{
+            "entity_name": <assign a name for the entity>,
+            "entity_type": <specify a short type name for the entity, such as 'company', 'location',...>
+        }}],
+        "relationships": [{{
+            "name": <assign a name for the relationship>,
+            "type": <specify a short type name for the relationship, such as 'sales_to', 'is_location_of',...>,
+            "entities": [<related entity name 1>, <related entity name 2>]
+        }}],
+        "terms": [{{
+            "term_name": <assign a name for the term>,
+            "term_type": <specify a short type name for the term, such as 'revenue', 'market_share',...>,
+            "similar_to": <list terms that are similar to this term>
+        }}]
+    }}
+    }}
+   """

backend/danswer/agent_search/shared_graph_utils/utils.py

@@ -0,0 +1,101 @@
+import ast
+import json
+import re
+from collections.abc import Sequence
+from datetime import datetime
+from datetime import timedelta
+from typing import Any
+
+from danswer.context.search.models import InferenceSection
+
+
+def normalize_whitespace(text: str) -> str:
+    """Normalize whitespace in text to single spaces and strip leading/trailing whitespace."""
+    import re
+
+    return re.sub(r"\s+", " ", text.strip())
+
+
+# Post-processing
+def format_docs(docs: Sequence[InferenceSection]) -> str:
+    return "\n\n".join(doc.combined_content for doc in docs)
+
+
+def clean_and_parse_list_string(json_string: str) -> list[dict]:
+    # Remove any prefixes/labels before the actual JSON content
+    json_string = re.sub(r"^.*?(?=\[)", "", json_string, flags=re.DOTALL)
+
+    # Remove markdown code block markers and any newline prefixes
+    cleaned_string = re.sub(r"```json\n|\n```", "", json_string)
+    cleaned_string = cleaned_string.replace("\\n", " ").replace("\n", " ")
+    cleaned_string = " ".join(cleaned_string.split())
+
+    # Try parsing with json.loads first, fall back to ast.literal_eval
+    try:
+        return json.loads(cleaned_string)
+    except json.JSONDecodeError:
+        try:
+            return ast.literal_eval(cleaned_string)
+        except (ValueError, SyntaxError) as e:
+            raise ValueError(f"Failed to parse JSON string: {cleaned_string}") from e
+
+
+def clean_and_parse_json_string(json_string: str) -> dict[str, Any]:
+    # Remove markdown code block markers and any newline prefixes
+    cleaned_string = re.sub(r"```json\n|\n```", "", json_string)
+    cleaned_string = cleaned_string.replace("\\n", " ").replace("\n", " ")
+    cleaned_string = " ".join(cleaned_string.split())
+    # Parse the cleaned string into a Python dictionary
+    return json.loads(cleaned_string)
+
+
+def format_entity_term_extraction(entity_term_extraction_dict: dict[str, Any]) -> str:
+    entities = entity_term_extraction_dict["entities"]
+    terms = entity_term_extraction_dict["terms"]
+    relationships = entity_term_extraction_dict["relationships"]
+
+    entity_strs = ["\nEntities:\n"]
+    for entity in entities:
+        entity_str = f"{entity['entity_name']} ({entity['entity_type']})"
+        entity_strs.append(entity_str)
+
+    entity_str = "\n - ".join(entity_strs)
+
+    relationship_strs = ["\n\nRelationships:\n"]
+    for relationship in relationships:
+        relationship_str = f"{relationship['name']} ({relationship['type']}): {relationship['entities']}"
+        relationship_strs.append(relationship_str)
+
+    relationship_str = "\n - ".join(relationship_strs)
+
+    term_strs = ["\n\nTerms:\n"]
+    for term in terms:
+        term_str = f"{term['term_name']} ({term['term_type']}): similar to {term['similar_to']}"
+        term_strs.append(term_str)
+
+    term_str = "\n - ".join(term_strs)
+
+    return "\n".join(entity_strs + relationship_strs + term_strs)
+
+
+def _format_time_delta(time: timedelta) -> str:
+    seconds_from_start = f"{((time).seconds):03d}"
+    microseconds_from_start = f"{((time).microseconds):06d}"
+    return f"{seconds_from_start}.{microseconds_from_start}"
+
+
+def generate_log_message(
+    message: str,
+    node_start_time: datetime,
+    graph_start_time: datetime | None = None,
+) -> str:
+    current_time = datetime.now()
+
+    if graph_start_time is not None:
+        graph_time_str = _format_time_delta(current_time - graph_start_time)
+    else:
+        graph_time_str = "N/A"
+
+    node_time_str = _format_time_delta(current_time - node_start_time)
+
+    return f"{graph_time_str} ({node_time_str} s): {message}"

backend/danswer/auth/noauth_user.py

@@ -23,7 +23,9 @@ def load_no_auth_user_preferences(store: KeyValueStore) -> UserPreferences:
         )
         return UserPreferences(**preferences_data)
     except KvKeyNotFoundError:
-        return UserPreferences(chosen_assistants=None, default_model=None)
+        return UserPreferences(
+            chosen_assistants=None, default_model=None, auto_scroll=True
+        )
 
 
 def fetch_no_auth_user(store: KeyValueStore) -> UserInfo:

backend/danswer/auth/users.py

@@ -87,6 +87,7 @@ from danswer.db.models import AccessToken
 from danswer.db.models import OAuthAccount
 from danswer.db.models import User
 from danswer.db.users import get_user_by_email
+from danswer.server.utils import BasicAuthenticationError
 from danswer.utils.logger import setup_logger
 from danswer.utils.telemetry import optional_telemetry
 from danswer.utils.telemetry import RecordType
@@ -99,11 +100,6 @@ from shared_configs.contextvars import CURRENT_TENANT_ID_CONTEXTVAR
 logger = setup_logger()
 
 
-class BasicAuthenticationError(HTTPException):
-    def __init__(self, detail: str):
-        super().__init__(status_code=status.HTTP_403_FORBIDDEN, detail=detail)
-
-
 def is_user_admin(user: User | None) -> bool:
     if AUTH_TYPE == AuthType.DISABLED:
         return True

backend/danswer/background/celery/apps/app_base.py

@@ -11,6 +11,7 @@ from celery.exceptions import WorkerShutdown
 from celery.states import READY_STATES
 from celery.utils.log import get_task_logger
 from celery.worker import strategy  # type: ignore
+from redis.lock import Lock as RedisLock
 from sentry_sdk.integrations.celery import CeleryIntegration
 from sqlalchemy import text
 from sqlalchemy.orm import Session
@@ -332,16 +333,16 @@ def on_worker_shutdown(sender: Any, **kwargs: Any) -> None:
         return
 
     logger.info("Releasing primary worker lock.")
-    lock = sender.primary_worker_lock
+    lock: RedisLock = sender.primary_worker_lock
     try:
         if lock.owned():
             try:
                 lock.release()
                 sender.primary_worker_lock = None
-            except Exception as e:
-                logger.error(f"Failed to release primary worker lock: {e}")
-    except Exception as e:
-        logger.error(f"Failed to check if primary worker lock is owned: {e}")
+            except Exception:
+                logger.exception("Failed to release primary worker lock")
+    except Exception:
+        logger.exception("Failed to check if primary worker lock is owned")
 
 
 def on_setup_logging(

backend/danswer/background/celery/apps/primary.py

@@ -11,6 +11,7 @@ from celery.signals import celeryd_init
 from celery.signals import worker_init
 from celery.signals import worker_ready
 from celery.signals import worker_shutdown
+from redis.lock import Lock as RedisLock
 
 import danswer.background.celery.apps.app_base as app_base
 from danswer.background.celery.apps.app_base import task_logger
@@ -38,7 +39,6 @@ from danswer.redis.redis_usergroup import RedisUserGroup
 from danswer.utils.logger import setup_logger
 from shared_configs.configs import MULTI_TENANT
 
-
 logger = setup_logger()
 
 celery_app = Celery(__name__)
@@ -116,9 +116,13 @@ def on_worker_init(sender: Any, **kwargs: Any) -> None:
     # it is planned to use this lock to enforce singleton behavior on the primary
     # worker, since the primary worker does redis cleanup on startup, but this isn't
     # implemented yet.
-    lock = r.lock(
+
+    # set thread_local=False since we don't control what thread the periodic task might
+    # reacquire the lock with
+    lock: RedisLock = r.lock(
         DanswerRedisLocks.PRIMARY_WORKER,
         timeout=CELERY_PRIMARY_WORKER_LOCK_TIMEOUT,
+        thread_local=False,
     )
 
     logger.info("Primary worker lock: Acquire starting.")
@@ -227,7 +231,7 @@ class HubPeriodicTask(bootsteps.StartStopStep):
             if not hasattr(worker, "primary_worker_lock"):
                 return
 
-            lock = worker.primary_worker_lock
+            lock: RedisLock = worker.primary_worker_lock
 
             r = get_redis_client(tenant_id=None)
 

backend/danswer/background/celery/tasks/beat_schedule.py

@@ -2,54 +2,55 @@ from datetime import timedelta
 from typing import Any
 
 from danswer.configs.constants import DanswerCeleryPriority
+from danswer.configs.constants import DanswerCeleryTask
 
 
 tasks_to_schedule = [
     {
         "name": "check-for-vespa-sync",
-        "task": "check_for_vespa_sync_task",
+        "task": DanswerCeleryTask.CHECK_FOR_VESPA_SYNC_TASK,
         "schedule": timedelta(seconds=20),
         "options": {"priority": DanswerCeleryPriority.HIGH},
     },
     {
         "name": "check-for-connector-deletion",
-        "task": "check_for_connector_deletion_task",
+        "task": DanswerCeleryTask.CHECK_FOR_CONNECTOR_DELETION,
         "schedule": timedelta(seconds=20),
         "options": {"priority": DanswerCeleryPriority.HIGH},
     },
     {
         "name": "check-for-indexing",
-        "task": "check_for_indexing",
+        "task": DanswerCeleryTask.CHECK_FOR_INDEXING,
         "schedule": timedelta(seconds=15),
         "options": {"priority": DanswerCeleryPriority.HIGH},
     },
     {
         "name": "check-for-prune",
-        "task": "check_for_pruning",
+        "task": DanswerCeleryTask.CHECK_FOR_PRUNING,
         "schedule": timedelta(seconds=15),
         "options": {"priority": DanswerCeleryPriority.HIGH},
     },
     {
         "name": "kombu-message-cleanup",
-        "task": "kombu_message_cleanup_task",
+        "task": DanswerCeleryTask.KOMBU_MESSAGE_CLEANUP_TASK,
         "schedule": timedelta(seconds=3600),
         "options": {"priority": DanswerCeleryPriority.LOWEST},
     },
     {
         "name": "monitor-vespa-sync",
-        "task": "monitor_vespa_sync",
+        "task": DanswerCeleryTask.MONITOR_VESPA_SYNC,
         "schedule": timedelta(seconds=5),
         "options": {"priority": DanswerCeleryPriority.HIGH},
     },
     {
         "name": "check-for-doc-permissions-sync",
-        "task": "check_for_doc_permissions_sync",
+        "task": DanswerCeleryTask.CHECK_FOR_DOC_PERMISSIONS_SYNC,
         "schedule": timedelta(seconds=30),
         "options": {"priority": DanswerCeleryPriority.HIGH},
     },
     {
         "name": "check-for-external-group-sync",
-        "task": "check_for_external_group_sync",
+        "task": DanswerCeleryTask.CHECK_FOR_EXTERNAL_GROUP_SYNC,
         "schedule": timedelta(seconds=20),
         "options": {"priority": DanswerCeleryPriority.HIGH},
     },

backend/danswer/background/celery/tasks/connector_deletion/tasks.py

@@ -5,13 +5,13 @@ from celery import Celery
 from celery import shared_task
 from celery import Task
 from celery.exceptions import SoftTimeLimitExceeded
-from redis import Redis
 from redis.lock import Lock as RedisLock
 from sqlalchemy.orm import Session
 
 from danswer.background.celery.apps.app_base import task_logger
 from danswer.configs.app_configs import JOB_TIMEOUT
 from danswer.configs.constants import CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT
+from danswer.configs.constants import DanswerCeleryTask
 from danswer.configs.constants import DanswerRedisLocks
 from danswer.db.connector_credential_pair import get_connector_credential_pair_from_id
 from danswer.db.connector_credential_pair import get_connector_credential_pairs
@@ -29,7 +29,7 @@ class TaskDependencyError(RuntimeError):
 
 
 @shared_task(
-    name="check_for_connector_deletion_task",
+    name=DanswerCeleryTask.CHECK_FOR_CONNECTOR_DELETION,
     soft_time_limit=JOB_TIMEOUT,
     trail=False,
     bind=True,
@@ -37,7 +37,7 @@ class TaskDependencyError(RuntimeError):
 def check_for_connector_deletion_task(self: Task, *, tenant_id: str | None) -> None:
     r = get_redis_client(tenant_id=tenant_id)
 
-    lock_beat = r.lock(
+    lock_beat: RedisLock = r.lock(
         DanswerRedisLocks.CHECK_CONNECTOR_DELETION_BEAT_LOCK,
         timeout=CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT,
     )
@@ -60,7 +60,7 @@ def check_for_connector_deletion_task(self: Task, *, tenant_id: str | None) -> N
                 redis_connector = RedisConnector(tenant_id, cc_pair_id)
                 try:
                     try_generate_document_cc_pair_cleanup_tasks(
-                        self.app, cc_pair_id, db_session, r, lock_beat, tenant_id
+                        self.app, cc_pair_id, db_session, lock_beat, tenant_id
                     )
                 except TaskDependencyError as e:
                     # this means we wanted to start deleting but dependent tasks were running
@@ -86,7 +86,6 @@ def try_generate_document_cc_pair_cleanup_tasks(
     app: Celery,
     cc_pair_id: int,
     db_session: Session,
-    r: Redis,
     lock_beat: RedisLock,
     tenant_id: str | None,
 ) -> int | None:

backend/danswer/background/celery/tasks/doc_permission_syncing/tasks.py

@@ -8,6 +8,7 @@ from celery import shared_task
 from celery import Task
 from celery.exceptions import SoftTimeLimitExceeded
 from redis import Redis
+from redis.lock import Lock as RedisLock
 
 from danswer.access.models import DocExternalAccess
 from danswer.background.celery.apps.app_base import task_logger
@@ -17,9 +18,11 @@ from danswer.configs.constants import CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT
 from danswer.configs.constants import DANSWER_REDIS_FUNCTION_LOCK_PREFIX
 from danswer.configs.constants import DanswerCeleryPriority
 from danswer.configs.constants import DanswerCeleryQueues
+from danswer.configs.constants import DanswerCeleryTask
 from danswer.configs.constants import DanswerRedisLocks
 from danswer.configs.constants import DocumentSource
 from danswer.db.connector_credential_pair import get_connector_credential_pair_from_id
+from danswer.db.document import upsert_document_by_connector_credential_pair
 from danswer.db.engine import get_session_with_tenant
 from danswer.db.enums import AccessType
 from danswer.db.enums import ConnectorCredentialPairStatus
@@ -27,7 +30,7 @@ from danswer.db.models import ConnectorCredentialPair
 from danswer.db.users import batch_add_ext_perm_user_if_not_exists
 from danswer.redis.redis_connector import RedisConnector
 from danswer.redis.redis_connector_doc_perm_sync import (
-    RedisConnectorPermissionSyncData,
+    RedisConnectorPermissionSyncPayload,
 )
 from danswer.redis.redis_pool import get_redis_client
 from danswer.utils.logger import doc_permission_sync_ctx
@@ -81,7 +84,7 @@ def _is_external_doc_permissions_sync_due(cc_pair: ConnectorCredentialPair) -> b
 
 
 @shared_task(
-    name="check_for_doc_permissions_sync",
+    name=DanswerCeleryTask.CHECK_FOR_DOC_PERMISSIONS_SYNC,
     soft_time_limit=JOB_TIMEOUT,
     bind=True,
 )
@@ -138,7 +141,7 @@ def try_creating_permissions_sync_task(
 
     LOCK_TIMEOUT = 30
 
-    lock = r.lock(
+    lock: RedisLock = r.lock(
         DANSWER_REDIS_FUNCTION_LOCK_PREFIX + "try_generate_permissions_sync_tasks",
         timeout=LOCK_TIMEOUT,
     )
@@ -162,8 +165,8 @@ def try_creating_permissions_sync_task(
 
         custom_task_id = f"{redis_connector.permissions.generator_task_key}_{uuid4()}"
 
-        app.send_task(
-            "connector_permission_sync_generator_task",
+        result = app.send_task(
+            DanswerCeleryTask.CONNECTOR_PERMISSION_SYNC_GENERATOR_TASK,
             kwargs=dict(
                 cc_pair_id=cc_pair_id,
                 tenant_id=tenant_id,
@@ -174,8 +177,8 @@ def try_creating_permissions_sync_task(
         )
 
         # set a basic fence to start
-        payload = RedisConnectorPermissionSyncData(
-            started=None,
+        payload = RedisConnectorPermissionSyncPayload(
+            started=None, celery_task_id=result.id
         )
 
         redis_connector.permissions.set_fence(payload)
@@ -190,7 +193,7 @@ def try_creating_permissions_sync_task(
 
 
 @shared_task(
-    name="connector_permission_sync_generator_task",
+    name=DanswerCeleryTask.CONNECTOR_PERMISSION_SYNC_GENERATOR_TASK,
     acks_late=False,
     soft_time_limit=JOB_TIMEOUT,
     track_started=True,
@@ -241,13 +244,17 @@ def connector_permission_sync_generator_task(
 
             doc_sync_func = DOC_PERMISSIONS_FUNC_MAP.get(source_type)
             if doc_sync_func is None:
-                raise ValueError(f"No doc sync func found for {source_type}")
+                raise ValueError(
+                    f"No doc sync func found for {source_type} with cc_pair={cc_pair_id}"
+                )
 
-            logger.info(f"Syncing docs for {source_type}")
+            logger.info(f"Syncing docs for {source_type} with cc_pair={cc_pair_id}")
 
-            payload = RedisConnectorPermissionSyncData(
-                started=datetime.now(timezone.utc),
-            )
+            payload = redis_connector.permissions.payload
+            if not payload:
+                raise ValueError(f"No fence payload found: cc_pair={cc_pair_id}")
+
+            payload.started = datetime.now(timezone.utc)
             redis_connector.permissions.set_fence(payload)
 
             document_external_accesses: list[DocExternalAccess] = doc_sync_func(cc_pair)
@@ -256,7 +263,12 @@ def connector_permission_sync_generator_task(
                 f"RedisConnector.permissions.generate_tasks starting. cc_pair={cc_pair_id}"
             )
             tasks_generated = redis_connector.permissions.generate_tasks(
-                self.app, lock, document_external_accesses, source_type
+                celery_app=self.app,
+                lock=lock,
+                new_permissions=document_external_accesses,
+                source_string=source_type,
+                connector_id=cc_pair.connector.id,
+                credential_id=cc_pair.credential.id,
             )
             if tasks_generated is None:
                 return None
@@ -281,7 +293,7 @@ def connector_permission_sync_generator_task(
 
 
 @shared_task(
-    name="update_external_document_permissions_task",
+    name=DanswerCeleryTask.UPDATE_EXTERNAL_DOCUMENT_PERMISSIONS_TASK,
     soft_time_limit=LIGHT_SOFT_TIME_LIMIT,
     time_limit=LIGHT_TIME_LIMIT,
     max_retries=DOCUMENT_PERMISSIONS_UPDATE_MAX_RETRIES,
@@ -292,6 +304,8 @@ def update_external_document_permissions_task(
     tenant_id: str | None,
     serialized_doc_external_access: dict,
     source_string: str,
+    connector_id: int,
+    credential_id: int,
 ) -> bool:
     document_external_access = DocExternalAccess.from_dict(
         serialized_doc_external_access
@@ -300,18 +314,28 @@ def update_external_document_permissions_task(
     external_access = document_external_access.external_access
     try:
         with get_session_with_tenant(tenant_id) as db_session:
-            # Then we build the update requests to update vespa
+            # Add the users to the DB if they don't exist
             batch_add_ext_perm_user_if_not_exists(
                 db_session=db_session,
                 emails=list(external_access.external_user_emails),
             )
-            upsert_document_external_perms(
+            # Then we upsert the document's external permissions in postgres
+            created_new_doc = upsert_document_external_perms(
                 db_session=db_session,
                 doc_id=doc_id,
                 external_access=external_access,
                 source_type=DocumentSource(source_string),
             )
 
+            if created_new_doc:
+                # If a new document was created, we associate it with the cc_pair
+                upsert_document_by_connector_credential_pair(
+                    db_session=db_session,
+                    connector_id=connector_id,
+                    credential_id=credential_id,
+                    document_ids=[doc_id],
+                )
+
             logger.debug(
                 f"Successfully synced postgres document permissions for {doc_id}"
             )

backend/danswer/background/celery/tasks/external_group_syncing/tasks.py

@@ -8,6 +8,7 @@ from celery import shared_task
 from celery import Task
 from celery.exceptions import SoftTimeLimitExceeded
 from redis import Redis
+from redis.lock import Lock as RedisLock
 
 from danswer.background.celery.apps.app_base import task_logger
 from danswer.configs.app_configs import JOB_TIMEOUT
@@ -16,6 +17,7 @@ from danswer.configs.constants import CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT
 from danswer.configs.constants import DANSWER_REDIS_FUNCTION_LOCK_PREFIX
 from danswer.configs.constants import DanswerCeleryPriority
 from danswer.configs.constants import DanswerCeleryQueues
+from danswer.configs.constants import DanswerCeleryTask
 from danswer.configs.constants import DanswerRedisLocks
 from danswer.db.connector import mark_cc_pair_as_external_group_synced
 from danswer.db.connector_credential_pair import get_connector_credential_pair_from_id
@@ -24,13 +26,20 @@ from danswer.db.enums import AccessType
 from danswer.db.enums import ConnectorCredentialPairStatus
 from danswer.db.models import ConnectorCredentialPair
 from danswer.redis.redis_connector import RedisConnector
+from danswer.redis.redis_connector_ext_group_sync import (
+    RedisConnectorExternalGroupSyncPayload,
+)
 from danswer.redis.redis_pool import get_redis_client
 from danswer.utils.logger import setup_logger
 from ee.danswer.db.connector_credential_pair import get_all_auto_sync_cc_pairs
+from ee.danswer.db.connector_credential_pair import get_cc_pairs_by_source
 from ee.danswer.db.external_perm import ExternalUserGroup
 from ee.danswer.db.external_perm import replace_user__ext_group_for_cc_pair
 from ee.danswer.external_permissions.sync_params import EXTERNAL_GROUP_SYNC_PERIODS
 from ee.danswer.external_permissions.sync_params import GROUP_PERMISSIONS_FUNC_MAP
+from ee.danswer.external_permissions.sync_params import (
+    GROUP_PERMISSIONS_IS_CC_PAIR_AGNOSTIC,
+)
 
 logger = setup_logger()
 
@@ -49,7 +58,7 @@ def _is_external_group_sync_due(cc_pair: ConnectorCredentialPair) -> bool:
     if cc_pair.access_type != AccessType.SYNC:
         return False
 
-    # skip pruning if not active
+    # skip external group sync if not active
     if cc_pair.status != ConnectorCredentialPairStatus.ACTIVE:
         return False
 
@@ -81,7 +90,7 @@ def _is_external_group_sync_due(cc_pair: ConnectorCredentialPair) -> bool:
 
 
 @shared_task(
-    name="check_for_external_group_sync",
+    name=DanswerCeleryTask.CHECK_FOR_EXTERNAL_GROUP_SYNC,
     soft_time_limit=JOB_TIMEOUT,
     bind=True,
 )
@@ -102,12 +111,28 @@ def check_for_external_group_sync(self: Task, *, tenant_id: str | None) -> None:
         with get_session_with_tenant(tenant_id) as db_session:
             cc_pairs = get_all_auto_sync_cc_pairs(db_session)
 
+            # We only want to sync one cc_pair per source type in
+            # GROUP_PERMISSIONS_IS_CC_PAIR_AGNOSTIC
+            for source in GROUP_PERMISSIONS_IS_CC_PAIR_AGNOSTIC:
+                # These are ordered by cc_pair id so the first one is the one we want
+                cc_pairs_to_dedupe = get_cc_pairs_by_source(
+                    db_session, source, only_sync=True
+                )
+                # We only want to sync one cc_pair per source type
+                # in GROUP_PERMISSIONS_IS_CC_PAIR_AGNOSTIC so we dedupe here
+                for cc_pair_to_remove in cc_pairs_to_dedupe[1:]:
+                    cc_pairs = [
+                        cc_pair
+                        for cc_pair in cc_pairs
+                        if cc_pair.id != cc_pair_to_remove.id
+                    ]
+
             for cc_pair in cc_pairs:
                 if _is_external_group_sync_due(cc_pair):
                     cc_pair_ids_to_sync.append(cc_pair.id)
 
         for cc_pair_id in cc_pair_ids_to_sync:
-            tasks_created = try_creating_permissions_sync_task(
+            tasks_created = try_creating_external_group_sync_task(
                 self.app, cc_pair_id, r, tenant_id
             )
             if not tasks_created:
@@ -125,7 +150,7 @@ def check_for_external_group_sync(self: Task, *, tenant_id: str | None) -> None:
             lock_beat.release()
 
 
-def try_creating_permissions_sync_task(
+def try_creating_external_group_sync_task(
     app: Celery,
     cc_pair_id: int,
     r: Redis,
@@ -156,8 +181,8 @@ def try_creating_permissions_sync_task(
 
         custom_task_id = f"{redis_connector.external_group_sync.taskset_key}_{uuid4()}"
 
-        _ = app.send_task(
-            "connector_external_group_sync_generator_task",
+        result = app.send_task(
+            DanswerCeleryTask.CONNECTOR_EXTERNAL_GROUP_SYNC_GENERATOR_TASK,
             kwargs=dict(
                 cc_pair_id=cc_pair_id,
                 tenant_id=tenant_id,
@@ -166,8 +191,13 @@ def try_creating_permissions_sync_task(
             task_id=custom_task_id,
             priority=DanswerCeleryPriority.HIGH,
         )
-        # set a basic fence to start
-        redis_connector.external_group_sync.set_fence(True)
+
+        payload = RedisConnectorExternalGroupSyncPayload(
+            started=datetime.now(timezone.utc),
+            celery_task_id=result.id,
+        )
+
+        redis_connector.external_group_sync.set_fence(payload)
 
     except Exception:
         task_logger.exception(
@@ -182,7 +212,7 @@ def try_creating_permissions_sync_task(
 
 
 @shared_task(
-    name="connector_external_group_sync_generator_task",
+    name=DanswerCeleryTask.CONNECTOR_EXTERNAL_GROUP_SYNC_GENERATOR_TASK,
     acks_late=False,
     soft_time_limit=JOB_TIMEOUT,
     track_started=True,
@@ -195,7 +225,7 @@ def connector_external_group_sync_generator_task(
     tenant_id: str | None,
 ) -> None:
     """
-    Permission sync task that handles document permission syncing for a given connector credential pair
+    Permission sync task that handles external group syncing for a given connector credential pair
     This task assumes that the task has already been properly fenced
     """
 
@@ -203,7 +233,7 @@ def connector_external_group_sync_generator_task(
 
     r = get_redis_client(tenant_id=tenant_id)
 
-    lock = r.lock(
+    lock: RedisLock = r.lock(
         DanswerRedisLocks.CONNECTOR_EXTERNAL_GROUP_SYNC_LOCK_PREFIX
         + f"_{redis_connector.id}",
         timeout=CELERY_EXTERNAL_GROUP_SYNC_LOCK_TIMEOUT,
@@ -228,9 +258,13 @@ def connector_external_group_sync_generator_task(
 
             ext_group_sync_func = GROUP_PERMISSIONS_FUNC_MAP.get(source_type)
             if ext_group_sync_func is None:
-                raise ValueError(f"No external group sync func found for {source_type}")
+                raise ValueError(
+                    f"No external group sync func found for {source_type} for cc_pair: {cc_pair_id}"
+                )
 
-            logger.info(f"Syncing docs for {source_type}")
+            logger.info(
+                f"Syncing external groups for {source_type} for cc_pair: {cc_pair_id}"
+            )
 
             external_user_groups: list[ExternalUserGroup] = ext_group_sync_func(cc_pair)
 
@@ -249,7 +283,6 @@ def connector_external_group_sync_generator_task(
             )
 
             mark_cc_pair_as_external_group_synced(db_session, cc_pair.id)
-
     except Exception as e:
         task_logger.exception(
             f"Failed to run external group sync: cc_pair={cc_pair_id}"
@@ -260,6 +293,6 @@ def connector_external_group_sync_generator_task(
         raise e
     finally:
         # we always want to clear the fence after the task is done or failed so it doesn't get stuck
-        redis_connector.external_group_sync.set_fence(False)
+        redis_connector.external_group_sync.set_fence(None)
         if lock.owned():
             lock.release()

backend/danswer/background/celery/tasks/indexing/tasks.py

@@ -23,13 +23,16 @@ from danswer.configs.constants import CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT
 from danswer.configs.constants import DANSWER_REDIS_FUNCTION_LOCK_PREFIX
 from danswer.configs.constants import DanswerCeleryPriority
 from danswer.configs.constants import DanswerCeleryQueues
+from danswer.configs.constants import DanswerCeleryTask
 from danswer.configs.constants import DanswerRedisLocks
 from danswer.configs.constants import DocumentSource
+from danswer.db.connector import mark_ccpair_with_indexing_trigger
 from danswer.db.connector_credential_pair import fetch_connector_credential_pairs
 from danswer.db.connector_credential_pair import get_connector_credential_pair_from_id
 from danswer.db.engine import get_db_current_time
 from danswer.db.engine import get_session_with_tenant
 from danswer.db.enums import ConnectorCredentialPairStatus
+from danswer.db.enums import IndexingMode
 from danswer.db.enums import IndexingStatus
 from danswer.db.enums import IndexModelStatus
 from danswer.db.index_attempt import create_index_attempt
@@ -37,12 +40,13 @@ from danswer.db.index_attempt import delete_index_attempt
 from danswer.db.index_attempt import get_all_index_attempts_by_status
 from danswer.db.index_attempt import get_index_attempt
 from danswer.db.index_attempt import get_last_attempt_for_cc_pair
+from danswer.db.index_attempt import mark_attempt_canceled
 from danswer.db.index_attempt import mark_attempt_failed
 from danswer.db.models import ConnectorCredentialPair
 from danswer.db.models import IndexAttempt
 from danswer.db.models import SearchSettings
+from danswer.db.search_settings import get_active_search_settings
 from danswer.db.search_settings import get_current_search_settings
-from danswer.db.search_settings import get_secondary_search_settings
 from danswer.db.swap_index import check_index_swap
 from danswer.indexing.indexing_heartbeat import IndexingHeartbeatInterface
 from danswer.natural_language_processing.search_nlp_models import EmbeddingModel
@@ -153,13 +157,13 @@ def get_unfenced_index_attempt_ids(db_session: Session, r: redis.Redis) -> list[
 
 
 @shared_task(
-    name="check_for_indexing",
+    name=DanswerCeleryTask.CHECK_FOR_INDEXING,
     soft_time_limit=300,
     bind=True,
 )
 def check_for_indexing(self: Task, *, tenant_id: str | None) -> int | None:
     tasks_created = 0
-
+    locked = False
     r = get_redis_client(tenant_id=tenant_id)
 
     lock_beat: RedisLock = r.lock(
@@ -172,6 +176,8 @@ def check_for_indexing(self: Task, *, tenant_id: str | None) -> int | None:
         if not lock_beat.acquire(blocking=False):
             return None
 
+        locked = True
+
         # check for search settings swap
         with get_session_with_tenant(tenant_id=tenant_id) as db_session:
             old_search_settings = check_index_swap(db_session=db_session)
@@ -205,17 +211,10 @@ def check_for_indexing(self: Task, *, tenant_id: str | None) -> int | None:
 
             redis_connector = RedisConnector(tenant_id, cc_pair_id)
             with get_session_with_tenant(tenant_id) as db_session:
-                # Get the primary search settings
-                primary_search_settings = get_current_search_settings(db_session)
-                search_settings = [primary_search_settings]
-
-                # Check for secondary search settings
-                secondary_search_settings = get_secondary_search_settings(db_session)
-                if secondary_search_settings is not None:
-                    # If secondary settings exist, add them to the list
-                    search_settings.append(secondary_search_settings)
-
-                for search_settings_instance in search_settings:
+                search_settings_list: list[SearchSettings] = get_active_search_settings(
+                    db_session
+                )
+                for search_settings_instance in search_settings_list:
                     redis_connector_index = redis_connector.new_index(
                         search_settings_instance.id
                     )
@@ -231,22 +230,46 @@ def check_for_indexing(self: Task, *, tenant_id: str | None) -> int | None:
                     last_attempt = get_last_attempt_for_cc_pair(
                         cc_pair.id, search_settings_instance.id, db_session
                     )
+
+                    search_settings_primary = False
+                    if search_settings_instance.id == search_settings_list[0].id:
+                        search_settings_primary = True
+
                     if not _should_index(
                         cc_pair=cc_pair,
                         last_index=last_attempt,
                         search_settings_instance=search_settings_instance,
-                        secondary_index_building=len(search_settings) > 1,
+                        search_settings_primary=search_settings_primary,
+                        secondary_index_building=len(search_settings_list) > 1,
                         db_session=db_session,
                     ):
                         continue
 
+                    reindex = False
+                    if search_settings_instance.id == search_settings_list[0].id:
+                        # the indexing trigger is only checked and cleared with the primary search settings
+                        if cc_pair.indexing_trigger is not None:
+                            if cc_pair.indexing_trigger == IndexingMode.REINDEX:
+                                reindex = True
+
+                            task_logger.info(
+                                f"Connector indexing manual trigger detected: "
+                                f"cc_pair={cc_pair.id} "
+                                f"search_settings={search_settings_instance.id} "
+                                f"indexing_mode={cc_pair.indexing_trigger}"
+                            )
+
+                            mark_ccpair_with_indexing_trigger(
+                                cc_pair.id, None, db_session
+                            )
+
                     # using a task queue and only allowing one task per cc_pair/search_setting
                     # prevents us from starving out certain attempts
                     attempt_id = try_creating_indexing_task(
                         self.app,
                         cc_pair,
                         search_settings_instance,
-                        False,
+                        reindex,
                         db_session,
                         r,
                         tenant_id,
@@ -256,7 +279,7 @@ def check_for_indexing(self: Task, *, tenant_id: str | None) -> int | None:
                             f"Connector indexing queued: "
                             f"index_attempt={attempt_id} "
                             f"cc_pair={cc_pair.id} "
-                            f"search_settings={search_settings_instance.id} "
+                            f"search_settings={search_settings_instance.id}"
                         )
                         tasks_created += 1
 
@@ -281,7 +304,6 @@ def check_for_indexing(self: Task, *, tenant_id: str | None) -> int | None:
                 mark_attempt_failed(
                     attempt.id, db_session, failure_reason=failure_reason
                 )
-
     except SoftTimeLimitExceeded:
         task_logger.info(
             "Soft time limit exceeded, task is being terminated gracefully."
@@ -289,13 +311,14 @@ def check_for_indexing(self: Task, *, tenant_id: str | None) -> int | None:
     except Exception:
         task_logger.exception(f"Unexpected exception: tenant={tenant_id}")
     finally:
-        if lock_beat.owned():
-            lock_beat.release()
-        else:
-            task_logger.error(
-                "check_for_indexing - Lock not owned on completion: "
-                f"tenant={tenant_id}"
-            )
+        if locked:
+            if lock_beat.owned():
+                lock_beat.release()
+            else:
+                task_logger.error(
+                    "check_for_indexing - Lock not owned on completion: "
+                    f"tenant={tenant_id}"
+                )
 
     return tasks_created
 
@@ -304,6 +327,7 @@ def _should_index(
     cc_pair: ConnectorCredentialPair,
     last_index: IndexAttempt | None,
     search_settings_instance: SearchSettings,
+    search_settings_primary: bool,
     secondary_index_building: bool,
     db_session: Session,
 ) -> bool:
@@ -368,6 +392,11 @@ def _should_index(
     ):
         return False
 
+    if search_settings_primary:
+        if cc_pair.indexing_trigger is not None:
+            # if a manual indexing trigger is on the cc pair, honor it for primary search settings
+            return True
+
     # if no attempt has ever occurred, we should index regardless of refresh_freq
     if not last_index:
         return True
@@ -458,7 +487,7 @@ def try_creating_indexing_task(
         # when the task is sent, we have yet to finish setting up the fence
         # therefore, the task must contain code that blocks until the fence is ready
         result = celery_app.send_task(
-            "connector_indexing_proxy_task",
+            DanswerCeleryTask.CONNECTOR_INDEXING_PROXY_TASK,
             kwargs=dict(
                 index_attempt_id=index_attempt_id,
                 cc_pair_id=cc_pair.id,
@@ -495,8 +524,14 @@ def try_creating_indexing_task(
     return index_attempt_id
 
 
-@shared_task(name="connector_indexing_proxy_task", acks_late=False, track_started=True)
+@shared_task(
+    name=DanswerCeleryTask.CONNECTOR_INDEXING_PROXY_TASK,
+    bind=True,
+    acks_late=False,
+    track_started=True,
+)
 def connector_indexing_proxy_task(
+    self: Task,
     index_attempt_id: int,
     cc_pair_id: int,
     search_settings_id: int,
@@ -509,6 +544,10 @@ def connector_indexing_proxy_task(
         f"cc_pair={cc_pair_id} "
         f"search_settings={search_settings_id}"
     )
+
+    if not self.request.id:
+        task_logger.error("self.request.id is None!")
+
     client = SimpleJobClient()
 
     job = client.submit(
@@ -537,25 +576,72 @@ def connector_indexing_proxy_task(
         f"search_settings={search_settings_id}"
     )
 
-    while True:
-        sleep(10)
+    redis_connector = RedisConnector(tenant_id, cc_pair_id)
+    redis_connector_index = redis_connector.new_index(search_settings_id)
 
-        # do nothing for ongoing jobs that haven't been stopped
-        if not job.done():
-            with get_session_with_tenant(tenant_id) as db_session:
-                index_attempt = get_index_attempt(
-                    db_session=db_session, index_attempt_id=index_attempt_id
+    while True:
+        sleep(5)
+
+        if self.request.id and redis_connector_index.terminating(self.request.id):
+            task_logger.warning(
+                "Indexing watchdog - termination signal detected: "
+                f"attempt={index_attempt_id} "
+                f"tenant={tenant_id} "
+                f"cc_pair={cc_pair_id} "
+                f"search_settings={search_settings_id}"
+            )
+
+            try:
+                with get_session_with_tenant(tenant_id) as db_session:
+                    mark_attempt_canceled(
+                        index_attempt_id,
+                        db_session,
+                        "Connector termination signal detected",
+                    )
+            finally:
+                # if the DB exceptions, we'll just get an unfriendly failure message
+                # in the UI instead of the cancellation message
+                logger.exception(
+                    "Indexing watchdog - transient exception marking index attempt as canceled: "
+                    f"attempt={index_attempt_id} "
+                    f"tenant={tenant_id} "
+                    f"cc_pair={cc_pair_id} "
+                    f"search_settings={search_settings_id}"
                 )
 
-                if not index_attempt:
-                    continue
+                job.cancel()
 
-                if not index_attempt.is_finished():
-                    continue
+            break
+
+        if not job.done():
+            # if the spawned task is still running, restart the check once again
+            # if the index attempt is not in a finished status
+            try:
+                with get_session_with_tenant(tenant_id) as db_session:
+                    index_attempt = get_index_attempt(
+                        db_session=db_session, index_attempt_id=index_attempt_id
+                    )
+
+                    if not index_attempt:
+                        continue
+
+                    if not index_attempt.is_finished():
+                        continue
+            except Exception:
+                # if the DB exceptioned, just restart the check.
+                # polling the index attempt status doesn't need to be strongly consistent
+                logger.exception(
+                    "Indexing watchdog - transient exception looking up index attempt: "
+                    f"attempt={index_attempt_id} "
+                    f"tenant={tenant_id} "
+                    f"cc_pair={cc_pair_id} "
+                    f"search_settings={search_settings_id}"
+                )
+                continue
 
         if job.status == "error":
             task_logger.error(
-                f"Indexing watchdog - spawned task exceptioned: "
+                "Indexing watchdog - spawned task exceptioned: "
                 f"attempt={index_attempt_id} "
                 f"tenant={tenant_id} "
                 f"cc_pair={cc_pair_id} "
@@ -703,9 +789,12 @@ def connector_indexing_task(
         )
         break
 
+    # set thread_local=False since we don't control what thread the indexing/pruning
+    # might run our callback with
     lock: RedisLock = r.lock(
         redis_connector_index.generator_lock_key,
         timeout=CELERY_INDEXING_LOCK_TIMEOUT,
+        thread_local=False,
     )
 
     acquired = lock.acquire(blocking=False)

backend/danswer/background/celery/tasks/periodic/tasks.py

@@ -13,12 +13,13 @@ from sqlalchemy.orm import Session
 
 from danswer.background.celery.apps.app_base import task_logger
 from danswer.configs.app_configs import JOB_TIMEOUT
+from danswer.configs.constants import DanswerCeleryTask
 from danswer.configs.constants import PostgresAdvisoryLocks
 from danswer.db.engine import get_session_with_tenant
 
 
 @shared_task(
-    name="kombu_message_cleanup_task",
+    name=DanswerCeleryTask.KOMBU_MESSAGE_CLEANUP_TASK,
     soft_time_limit=JOB_TIMEOUT,
     bind=True,
     base=AbortableTask,

backend/danswer/background/celery/tasks/pruning/tasks.py

@@ -8,6 +8,7 @@ from celery import shared_task
 from celery import Task
 from celery.exceptions import SoftTimeLimitExceeded
 from redis import Redis
+from redis.lock import Lock as RedisLock
 from sqlalchemy.orm import Session
 
 from danswer.background.celery.apps.app_base import task_logger
@@ -20,6 +21,7 @@ from danswer.configs.constants import CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT
 from danswer.configs.constants import DANSWER_REDIS_FUNCTION_LOCK_PREFIX
 from danswer.configs.constants import DanswerCeleryPriority
 from danswer.configs.constants import DanswerCeleryQueues
+from danswer.configs.constants import DanswerCeleryTask
 from danswer.configs.constants import DanswerRedisLocks
 from danswer.connectors.factory import instantiate_connector
 from danswer.connectors.models import InputType
@@ -75,7 +77,7 @@ def _is_pruning_due(cc_pair: ConnectorCredentialPair) -> bool:
 
 
 @shared_task(
-    name="check_for_pruning",
+    name=DanswerCeleryTask.CHECK_FOR_PRUNING,
     soft_time_limit=JOB_TIMEOUT,
     bind=True,
 )
@@ -184,7 +186,7 @@ def try_creating_prune_generator_task(
         custom_task_id = f"{redis_connector.prune.generator_task_key}_{uuid4()}"
 
         celery_app.send_task(
-            "connector_pruning_generator_task",
+            DanswerCeleryTask.CONNECTOR_PRUNING_GENERATOR_TASK,
             kwargs=dict(
                 cc_pair_id=cc_pair.id,
                 connector_id=cc_pair.connector_id,
@@ -209,7 +211,7 @@ def try_creating_prune_generator_task(
 
 
 @shared_task(
-    name="connector_pruning_generator_task",
+    name=DanswerCeleryTask.CONNECTOR_PRUNING_GENERATOR_TASK,
     acks_late=False,
     soft_time_limit=JOB_TIMEOUT,
     track_started=True,
@@ -238,9 +240,12 @@ def connector_pruning_generator_task(
 
     r = get_redis_client(tenant_id=tenant_id)
 
-    lock = r.lock(
+    # set thread_local=False since we don't control what thread the indexing/pruning
+    # might run our callback with
+    lock: RedisLock = r.lock(
         DanswerRedisLocks.PRUNING_LOCK_PREFIX + f"_{redis_connector.id}",
         timeout=CELERY_PRUNING_LOCK_TIMEOUT,
+        thread_local=False,
     )
 
     acquired = lock.acquire(blocking=False)

backend/danswer/background/celery/tasks/shared/tasks.py

@@ -9,6 +9,7 @@ from tenacity import RetryError
 from danswer.access.access import get_access_for_document
 from danswer.background.celery.apps.app_base import task_logger
 from danswer.background.celery.tasks.shared.RetryDocumentIndex import RetryDocumentIndex
+from danswer.configs.constants import DanswerCeleryTask
 from danswer.db.document import delete_document_by_connector_credential_pair__no_commit
 from danswer.db.document import delete_documents_complete__no_commit
 from danswer.db.document import get_document
@@ -31,7 +32,7 @@ LIGHT_TIME_LIMIT = LIGHT_SOFT_TIME_LIMIT + 15
 
 
 @shared_task(
-    name="document_by_cc_pair_cleanup_task",
+    name=DanswerCeleryTask.DOCUMENT_BY_CC_PAIR_CLEANUP_TASK,
     soft_time_limit=LIGHT_SOFT_TIME_LIMIT,
     time_limit=LIGHT_TIME_LIMIT,
     max_retries=DOCUMENT_BY_CC_PAIR_CLEANUP_MAX_RETRIES,

backend/danswer/background/celery/tasks/vespa/tasks.py

@@ -25,6 +25,7 @@ from danswer.background.celery.tasks.shared.tasks import LIGHT_TIME_LIMIT
 from danswer.configs.app_configs import JOB_TIMEOUT
 from danswer.configs.constants import CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT
 from danswer.configs.constants import DanswerCeleryQueues
+from danswer.configs.constants import DanswerCeleryTask
 from danswer.configs.constants import DanswerRedisLocks
 from danswer.db.connector import fetch_connector_by_id
 from danswer.db.connector import mark_cc_pair_as_permissions_synced
@@ -46,6 +47,7 @@ from danswer.db.document_set import fetch_document_sets_for_document
 from danswer.db.document_set import get_document_set_by_id
 from danswer.db.document_set import mark_document_set_as_synced
 from danswer.db.engine import get_session_with_tenant
+from danswer.db.enums import IndexingStatus
 from danswer.db.index_attempt import delete_index_attempts
 from danswer.db.index_attempt import get_index_attempt
 from danswer.db.index_attempt import mark_attempt_failed
@@ -58,7 +60,7 @@ from danswer.redis.redis_connector_credential_pair import RedisConnectorCredenti
 from danswer.redis.redis_connector_delete import RedisConnectorDelete
 from danswer.redis.redis_connector_doc_perm_sync import RedisConnectorPermissionSync
 from danswer.redis.redis_connector_doc_perm_sync import (
-    RedisConnectorPermissionSyncData,
+    RedisConnectorPermissionSyncPayload,
 )
 from danswer.redis.redis_connector_index import RedisConnectorIndex
 from danswer.redis.redis_connector_prune import RedisConnectorPrune
@@ -79,7 +81,7 @@ logger = setup_logger()
 # celery auto associates tasks created inside another task,
 # which bloats the result metadata considerably. trail=False prevents this.
 @shared_task(
-    name="check_for_vespa_sync_task",
+    name=DanswerCeleryTask.CHECK_FOR_VESPA_SYNC_TASK,
     soft_time_limit=JOB_TIMEOUT,
     trail=False,
     bind=True,
@@ -588,7 +590,7 @@ def monitor_ccpair_permissions_taskset(
     if remaining > 0:
         return
 
-    payload: RedisConnectorPermissionSyncData | None = (
+    payload: RedisConnectorPermissionSyncPayload | None = (
         redis_connector.permissions.payload
     )
     start_time: datetime | None = payload.started if payload else None
@@ -596,9 +598,7 @@ def monitor_ccpair_permissions_taskset(
     mark_cc_pair_as_permissions_synced(db_session, int(cc_pair_id), start_time)
     task_logger.info(f"Successfully synced permissions for cc_pair={cc_pair_id}")
 
-    redis_connector.permissions.taskset_clear()
-    redis_connector.permissions.generator_clear()
-    redis_connector.permissions.set_fence(None)
+    redis_connector.permissions.reset()
 
 
 def monitor_ccpair_indexing_taskset(
@@ -655,34 +655,42 @@ def monitor_ccpair_indexing_taskset(
     # outer = result.state in READY state
     status_int = redis_connector_index.get_completion()
     if status_int is None:  # inner signal not set ... possible error
-        result_state = result.state
+        task_state = result.state
         if (
-            result_state in READY_STATES
+            task_state in READY_STATES
         ):  # outer signal in terminal state ... possible error
             # Now double check!
             if redis_connector_index.get_completion() is None:
                 # inner signal still not set (and cannot change when outer result_state is READY)
                 # Task is finished but generator complete isn't set.
                 # We have a problem! Worker may have crashed.
+                task_result = str(result.result)
+                task_traceback = str(result.traceback)
 
                 msg = (
                     f"Connector indexing aborted or exceptioned: "
                     f"attempt={payload.index_attempt_id} "
                     f"celery_task={payload.celery_task_id} "
-                    f"result_state={result_state} "
                     f"cc_pair={cc_pair_id} "
                     f"search_settings={search_settings_id} "
-                    f"elapsed_submitted={elapsed_submitted.total_seconds():.2f}"
+                    f"elapsed_submitted={elapsed_submitted.total_seconds():.2f} "
+                    f"result.state={task_state} "
+                    f"result.result={task_result} "
+                    f"result.traceback={task_traceback}"
                 )
                 task_logger.warning(msg)
 
                 index_attempt = get_index_attempt(db_session, payload.index_attempt_id)
                 if index_attempt:
-                    mark_attempt_failed(
-                        index_attempt_id=payload.index_attempt_id,
-                        db_session=db_session,
-                        failure_reason=msg,
-                    )
+                    if (
+                        index_attempt.status != IndexingStatus.CANCELED
+                        and index_attempt.status != IndexingStatus.FAILED
+                    ):
+                        mark_attempt_failed(
+                            index_attempt_id=payload.index_attempt_id,
+                            db_session=db_session,
+                            failure_reason=msg,
+                        )
 
                 redis_connector_index.reset()
         return
@@ -692,6 +700,7 @@ def monitor_ccpair_indexing_taskset(
     task_logger.info(
         f"Connector indexing finished: cc_pair={cc_pair_id} "
         f"search_settings={search_settings_id} "
+        f"progress={progress} "
         f"status={status_enum.name} "
         f"elapsed_submitted={elapsed_submitted.total_seconds():.2f}"
     )
@@ -699,7 +708,7 @@ def monitor_ccpair_indexing_taskset(
     redis_connector_index.reset()
 
 
-@shared_task(name="monitor_vespa_sync", soft_time_limit=300, bind=True)
+@shared_task(name=DanswerCeleryTask.MONITOR_VESPA_SYNC, soft_time_limit=300, bind=True)
 def monitor_vespa_sync(self: Task, tenant_id: str | None) -> bool:
     """This is a celery beat task that monitors and finalizes metadata sync tasksets.
     It scans for fence values and then gets the counts of any associated tasksets.
@@ -724,7 +733,7 @@ def monitor_vespa_sync(self: Task, tenant_id: str | None) -> bool:
 
         # print current queue lengths
         r_celery = self.app.broker_connection().channel().client  # type: ignore
-        n_celery = celery_get_queue_length("celery", r)
+        n_celery = celery_get_queue_length("celery", r_celery)
         n_indexing = celery_get_queue_length(
             DanswerCeleryQueues.CONNECTOR_INDEXING, r_celery
         )
@@ -810,7 +819,7 @@ def monitor_vespa_sync(self: Task, tenant_id: str | None) -> bool:
 
 
 @shared_task(
-    name="vespa_metadata_sync_task",
+    name=DanswerCeleryTask.VESPA_METADATA_SYNC_TASK,
     bind=True,
     soft_time_limit=LIGHT_SOFT_TIME_LIMIT,
     time_limit=LIGHT_TIME_LIMIT,

backend/danswer/background/celery/versioned_apps/beat.py

@@ -1,6 +1,8 @@
 """Factory stub for running celery worker / celery beat."""
+from celery import Celery
+
 from danswer.background.celery.apps.beat import celery_app
 from danswer.utils.variable_functionality import set_is_ee_based_on_env_variable
 
 set_is_ee_based_on_env_variable()
-app = celery_app
+app: Celery = celery_app

backend/danswer/background/celery/versioned_apps/primary.py

@@ -1,8 +1,10 @@
 """Factory stub for running celery worker / celery beat."""
+from celery import Celery
+
 from danswer.utils.variable_functionality import fetch_versioned_implementation
 from danswer.utils.variable_functionality import set_is_ee_based_on_env_variable
 
 set_is_ee_based_on_env_variable()
-app = fetch_versioned_implementation(
+app: Celery = fetch_versioned_implementation(
     "danswer.background.celery.apps.primary", "celery_app"
 )

backend/danswer/background/indexing/run_indexing.py

@@ -19,6 +19,7 @@ from danswer.db.connector_credential_pair import get_last_successful_attempt_tim
 from danswer.db.connector_credential_pair import update_connector_credential_pair
 from danswer.db.engine import get_session_with_tenant
 from danswer.db.enums import ConnectorCredentialPairStatus
+from danswer.db.index_attempt import mark_attempt_canceled
 from danswer.db.index_attempt import mark_attempt_failed
 from danswer.db.index_attempt import mark_attempt_partially_succeeded
 from danswer.db.index_attempt import mark_attempt_succeeded
@@ -87,6 +88,10 @@ def _get_connector_runner(
     )
 
 
+class ConnectorStopSignal(Exception):
+    """A custom exception used to signal a stop in processing."""
+
+
 def _run_indexing(
     db_session: Session,
     index_attempt: IndexAttempt,
@@ -208,9 +213,7 @@ def _run_indexing(
                 # contents still need to be initially pulled.
                 if callback:
                     if callback.should_stop():
-                        raise RuntimeError(
-                            "_run_indexing: Connector stop signal detected"
-                        )
+                        raise ConnectorStopSignal("Connector stop signal detected")
 
                 # TODO: should we move this into the above callback instead?
                 db_session.refresh(db_cc_pair)
@@ -304,26 +307,16 @@ def _run_indexing(
                 )
         except Exception as e:
             logger.exception(
-                f"Connector run ran into exception after elapsed time: {time.time() - start_time} seconds"
+                f"Connector run exceptioned after elapsed time: {time.time() - start_time} seconds"
             )
-            # Only mark the attempt as a complete failure if this is the first indexing window.
-            # Otherwise, some progress was made - the next run will not start from the beginning.
-            # In this case, it is not accurate to mark it as a failure. When the next run begins,
-            # if that fails immediately, it will be marked as a failure.
-            #
-            # NOTE: if the connector is manually disabled, we should mark it as a failure regardless
-            # to give better clarity in the UI, as the next run will never happen.
-            if (
-                ind == 0
-                or not db_cc_pair.status.is_active()
-                or index_attempt.status != IndexingStatus.IN_PROGRESS
-            ):
-                mark_attempt_failed(
+
+            if isinstance(e, ConnectorStopSignal):
+                mark_attempt_canceled(
                     index_attempt.id,
                     db_session,
-                    failure_reason=str(e),
-                    full_exception_trace=traceback.format_exc(),
+                    reason=str(e),
                 )
+
                 if is_primary:
                     update_connector_credential_pair(
                         db_session=db_session,
@@ -335,6 +328,37 @@ def _run_indexing(
                 if INDEXING_TRACER_INTERVAL > 0:
                     tracer.stop()
                 raise e
+            else:
+                # Only mark the attempt as a complete failure if this is the first indexing window.
+                # Otherwise, some progress was made - the next run will not start from the beginning.
+                # In this case, it is not accurate to mark it as a failure. When the next run begins,
+                # if that fails immediately, it will be marked as a failure.
+                #
+                # NOTE: if the connector is manually disabled, we should mark it as a failure regardless
+                # to give better clarity in the UI, as the next run will never happen.
+                if (
+                    ind == 0
+                    or not db_cc_pair.status.is_active()
+                    or index_attempt.status != IndexingStatus.IN_PROGRESS
+                ):
+                    mark_attempt_failed(
+                        index_attempt.id,
+                        db_session,
+                        failure_reason=str(e),
+                        full_exception_trace=traceback.format_exc(),
+                    )
+
+                    if is_primary:
+                        update_connector_credential_pair(
+                            db_session=db_session,
+                            connector_id=db_connector.id,
+                            credential_id=db_credential.id,
+                            net_docs=net_doc_change,
+                        )
+
+                    if INDEXING_TRACER_INTERVAL > 0:
+                        tracer.stop()
+                    raise e
 
             # break => similar to success case. As mentioned above, if the next run fails for the same
             # reason it will then be marked as a failure

backend/danswer/chat/chat_utils.py

@@ -2,20 +2,79 @@ import re
 from typing import cast
 from uuid import UUID
 
+from fastapi import HTTPException
 from fastapi.datastructures import Headers
 from sqlalchemy.orm import Session
 
+from danswer.auth.users import is_user_admin
 from danswer.chat.models import CitationInfo
 from danswer.chat.models import LlmDoc
+from danswer.chat.models import PersonaOverrideConfig
+from danswer.chat.models import ThreadMessage
+from danswer.configs.constants import DEFAULT_PERSONA_ID
+from danswer.configs.constants import MessageType
 from danswer.context.search.models import InferenceSection
+from danswer.context.search.models import RerankingDetails
+from danswer.context.search.models import RetrievalDetails
+from danswer.db.chat import create_chat_session
 from danswer.db.chat import get_chat_messages_by_session
+from danswer.db.llm import fetch_existing_doc_sets
+from danswer.db.llm import fetch_existing_tools
 from danswer.db.models import ChatMessage
+from danswer.db.models import Persona
+from danswer.db.models import Prompt
+from danswer.db.models import Tool
+from danswer.db.models import User
+from danswer.db.persona import get_prompts_by_ids
 from danswer.llm.answering.models import PreviousMessage
+from danswer.natural_language_processing.utils import BaseTokenizer
+from danswer.server.query_and_chat.models import CreateChatMessageRequest
+from danswer.tools.tool_implementations.custom.custom_tool import (
+    build_custom_tools_from_openapi_schema_and_headers,
+)
 from danswer.utils.logger import setup_logger
 
 logger = setup_logger()
 
 
+def prepare_chat_message_request(
+    message_text: str,
+    user: User | None,
+    persona_id: int | None,
+    # Does the question need to have a persona override
+    persona_override_config: PersonaOverrideConfig | None,
+    prompt: Prompt | None,
+    message_ts_to_respond_to: str | None,
+    retrieval_details: RetrievalDetails | None,
+    rerank_settings: RerankingDetails | None,
+    db_session: Session,
+) -> CreateChatMessageRequest:
+    # Typically used for one shot flows like SlackBot or non-chat API endpoint use cases
+    new_chat_session = create_chat_session(
+        db_session=db_session,
+        description=None,
+        user_id=user.id if user else None,
+        # If using an override, this id will be ignored later on
+        persona_id=persona_id or DEFAULT_PERSONA_ID,
+        danswerbot_flow=True,
+        slack_thread_id=message_ts_to_respond_to,
+    )
+
+    return CreateChatMessageRequest(
+        chat_session_id=new_chat_session.id,
+        parent_message_id=None,  # It's a standalone chat session each time
+        message=message_text,
+        file_descriptors=[],  # Currently SlackBot/answer api do not support files in the context
+        prompt_id=prompt.id if prompt else None,
+        # Can always override the persona for the single query, if it's a normal persona
+        # then it will be treated the same
+        persona_override_config=persona_override_config,
+        search_doc_ids=None,
+        retrieval_options=retrieval_details,
+        rerank_settings=rerank_settings,
+    )
+
+
 def llm_doc_from_inference_section(inference_section: InferenceSection) -> LlmDoc:
     return LlmDoc(
         document_id=inference_section.center_chunk.document_id,
@@ -31,9 +90,49 @@ def llm_doc_from_inference_section(inference_section: InferenceSection) -> LlmDo
         if inference_section.center_chunk.source_links
         else None,
         source_links=inference_section.center_chunk.source_links,
+        match_highlights=inference_section.center_chunk.match_highlights,
     )
 
 
+def combine_message_thread(
+    messages: list[ThreadMessage],
+    max_tokens: int | None,
+    llm_tokenizer: BaseTokenizer,
+) -> str:
+    """Used to create a single combined message context from threads"""
+    if not messages:
+        return ""
+
+    message_strs: list[str] = []
+    total_token_count = 0
+
+    for message in reversed(messages):
+        if message.role == MessageType.USER:
+            role_str = message.role.value.upper()
+            if message.sender:
+                role_str += " " + message.sender
+            else:
+                # Since other messages might have the user identifying information
+                # better to use Unknown for symmetry
+                role_str += " Unknown"
+        else:
+            role_str = message.role.value.upper()
+
+        msg_str = f"{role_str}:\n{message.message}"
+        message_token_count = len(llm_tokenizer.encode(msg_str))
+
+        if (
+            max_tokens is not None
+            and total_token_count + message_token_count > max_tokens
+        ):
+            break
+
+        message_strs.insert(0, msg_str)
+        total_token_count += message_token_count
+
+    return "\n\n".join(message_strs)
+
+
 def create_chat_chain(
     chat_session_id: UUID,
     db_session: Session,
@@ -196,3 +295,71 @@ def extract_headers(
             if lowercase_key in headers:
                 extracted_headers[lowercase_key] = headers[lowercase_key]
     return extracted_headers
+
+
+def create_temporary_persona(
+    persona_config: PersonaOverrideConfig, db_session: Session, user: User | None = None
+) -> Persona:
+    if not is_user_admin(user):
+        raise HTTPException(
+            status_code=403,
+            detail="User is not authorized to create a persona in one shot queries",
+        )
+
+    """Create a temporary Persona object from the provided configuration."""
+    persona = Persona(
+        name=persona_config.name,
+        description=persona_config.description,
+        num_chunks=persona_config.num_chunks,
+        llm_relevance_filter=persona_config.llm_relevance_filter,
+        llm_filter_extraction=persona_config.llm_filter_extraction,
+        recency_bias=persona_config.recency_bias,
+        llm_model_provider_override=persona_config.llm_model_provider_override,
+        llm_model_version_override=persona_config.llm_model_version_override,
+    )
+
+    if persona_config.prompts:
+        persona.prompts = [
+            Prompt(
+                name=p.name,
+                description=p.description,
+                system_prompt=p.system_prompt,
+                task_prompt=p.task_prompt,
+                include_citations=p.include_citations,
+                datetime_aware=p.datetime_aware,
+            )
+            for p in persona_config.prompts
+        ]
+    elif persona_config.prompt_ids:
+        persona.prompts = get_prompts_by_ids(
+            db_session=db_session, prompt_ids=persona_config.prompt_ids
+        )
+
+    persona.tools = []
+    if persona_config.custom_tools_openapi:
+        for schema in persona_config.custom_tools_openapi:
+            tools = cast(
+                list[Tool],
+                build_custom_tools_from_openapi_schema_and_headers(schema),
+            )
+            persona.tools.extend(tools)
+
+    if persona_config.tools:
+        tool_ids = [tool.id for tool in persona_config.tools]
+        persona.tools.extend(
+            fetch_existing_tools(db_session=db_session, tool_ids=tool_ids)
+        )
+
+    if persona_config.tool_ids:
+        persona.tools.extend(
+            fetch_existing_tools(
+                db_session=db_session, tool_ids=persona_config.tool_ids
+            )
+        )
+
+    fetched_docs = fetch_existing_doc_sets(
+        db_session=db_session, doc_ids=persona_config.document_set_ids
+    )
+    persona.document_sets = fetched_docs
+
+    return persona

backend/danswer/chat/models.py

@@ -4,12 +4,14 @@ from enum import Enum
 from typing import Any
 
 from pydantic import BaseModel
+from pydantic import Field
 
 from danswer.configs.constants import DocumentSource
+from danswer.configs.constants import MessageType
 from danswer.context.search.enums import QueryFlow
+from danswer.context.search.enums import RecencyBiasSetting
 from danswer.context.search.enums import SearchType
 from danswer.context.search.models import RetrievalDocs
-from danswer.context.search.models import SearchResponse
 from danswer.tools.tool_implementations.custom.base_tool_types import ToolResultType
 
 
@@ -25,6 +27,7 @@ class LlmDoc(BaseModel):
     updated_at: datetime | None
     link: str | None
     source_links: dict[int, str] | None
+    match_highlights: list[str] | None
 
 
 # First chunk of info for streaming QA
@@ -117,20 +120,6 @@ class StreamingError(BaseModel):
     stack_trace: str | None = None
 
 
-class DanswerQuote(BaseModel):
-    # This is during inference so everything is a string by this point
-    quote: str
-    document_id: str
-    link: str | None
-    source_type: str
-    semantic_identifier: str
-    blurb: str
-
-
-class DanswerQuotes(BaseModel):
-    quotes: list[DanswerQuote]
-
-
 class DanswerContext(BaseModel):
     content: str
     document_id: str
@@ -146,14 +135,20 @@ class DanswerAnswer(BaseModel):
     answer: str | None
 
 
-class QAResponse(SearchResponse, DanswerAnswer):
-    quotes: list[DanswerQuote] | None
-    contexts: list[DanswerContexts] | None
-    predicted_flow: QueryFlow
-    predicted_search: SearchType
-    eval_res_valid: bool | None = None
+class ThreadMessage(BaseModel):
+    message: str
+    sender: str | None = None
+    role: MessageType = MessageType.USER
+
+
+class ChatDanswerBotResponse(BaseModel):
+    answer: str | None = None
+    citations: list[CitationInfo] | None = None
+    docs: QADocsResponse | None = None
     llm_selected_doc_indices: list[int] | None = None
     error_msg: str | None = None
+    chat_message_id: int | None = None
+    answer_valid: bool = True  # Reflexion result, default True if Reflexion not run
 
 
 class FileChatDisplay(BaseModel):
@@ -165,9 +160,41 @@ class CustomToolResponse(BaseModel):
     tool_name: str
 
 
+class ToolConfig(BaseModel):
+    id: int
+
+
+class PromptOverrideConfig(BaseModel):
+    name: str
+    description: str = ""
+    system_prompt: str
+    task_prompt: str = ""
+    include_citations: bool = True
+    datetime_aware: bool = True
+
+
+class PersonaOverrideConfig(BaseModel):
+    name: str
+    description: str
+    search_type: SearchType = SearchType.SEMANTIC
+    num_chunks: float | None = None
+    llm_relevance_filter: bool = False
+    llm_filter_extraction: bool = False
+    recency_bias: RecencyBiasSetting = RecencyBiasSetting.AUTO
+    llm_model_provider_override: str | None = None
+    llm_model_version_override: str | None = None
+
+    prompts: list[PromptOverrideConfig] = Field(default_factory=list)
+    prompt_ids: list[int] = Field(default_factory=list)
+
+    document_set_ids: list[int] = Field(default_factory=list)
+    tools: list[ToolConfig] = Field(default_factory=list)
+    tool_ids: list[int] = Field(default_factory=list)
+    custom_tools_openapi: list[dict[str, Any]] = Field(default_factory=list)
+
+
 AnswerQuestionPossibleReturn = (
     DanswerAnswerPiece
-    | DanswerQuotes
     | CitationInfo
     | DanswerContexts
     | FileChatDisplay

backend/danswer/chat/process_message.py

@@ -7,10 +7,13 @@ from typing import cast
 from sqlalchemy.orm import Session
 
 from danswer.chat.chat_utils import create_chat_chain
+from danswer.chat.chat_utils import create_temporary_persona
 from danswer.chat.models import AllCitations
+from danswer.chat.models import ChatDanswerBotResponse
 from danswer.chat.models import CitationInfo
 from danswer.chat.models import CustomToolResponse
 from danswer.chat.models import DanswerAnswerPiece
+from danswer.chat.models import DanswerContexts
 from danswer.chat.models import FileChatDisplay
 from danswer.chat.models import FinalUsedContextDocsResponse
 from danswer.chat.models import LLMRelevanceFilterResponse
@@ -102,6 +105,7 @@ from danswer.tools.tool_implementations.internet_search.internet_search_tool imp
 from danswer.tools.tool_implementations.search.search_tool import (
     FINAL_CONTEXT_DOCUMENTS_ID,
 )
+from danswer.tools.tool_implementations.search.search_tool import SEARCH_DOC_CONTENT_ID
 from danswer.tools.tool_implementations.search.search_tool import (
     SEARCH_RESPONSE_SUMMARY_ID,
 )
@@ -113,8 +117,10 @@ from danswer.tools.tool_implementations.search.search_tool import (
 from danswer.tools.tool_runner import ToolCallFinalResult
 from danswer.utils.logger import setup_logger
 from danswer.utils.long_term_log import LongTermLogger
+from danswer.utils.timing import log_function_time
 from danswer.utils.timing import log_generator_function_time
 
+
 logger = setup_logger()
 
 
@@ -256,6 +262,7 @@ def _get_force_search_settings(
 ChatPacket = (
     StreamingError
     | QADocsResponse
+    | DanswerContexts
     | LLMRelevanceFilterResponse
     | FinalUsedContextDocsResponse
     | ChatMessageDetail
@@ -286,6 +293,8 @@ def stream_chat_message_objects(
     custom_tool_additional_headers: dict[str, str] | None = None,
     is_connected: Callable[[], bool] | None = None,
     enforce_chat_session_id_for_search_docs: bool = True,
+    bypass_acl: bool = False,
+    include_contexts: bool = False,
 ) -> ChatPacketStream:
     """Streams in order:
     1. [conditional] Retrieved documents if a search needs to be run
@@ -322,17 +331,31 @@ def stream_chat_message_objects(
             metadata={"user_id": str(user_id), "chat_session_id": str(chat_session_id)}
         )
 
-        # use alternate persona if alternative assistant id is passed in
         if alternate_assistant_id is not None:
+            # Allows users to specify a temporary persona (assistant) in the chat session
+            # this takes highest priority since it's user specified
             persona = get_persona_by_id(
                 alternate_assistant_id,
                 user=user,
                 db_session=db_session,
                 is_for_edit=False,
             )
+        elif new_msg_req.persona_override_config:
+            # Certain endpoints allow users to specify arbitrary persona settings
+            # this should never conflict with the alternate_assistant_id
+            persona = persona = create_temporary_persona(
+                db_session=db_session,
+                persona_config=new_msg_req.persona_override_config,
+                user=user,
+            )
         else:
             persona = chat_session.persona
 
+        if not persona:
+            raise RuntimeError("No persona specified or found for chat session")
+
+        # If a prompt override is specified via the API, use that with highest priority
+        # but for saving it, we are just mapping it to an existing prompt
         prompt_id = new_msg_req.prompt_id
         if prompt_id is None and persona.prompts:
             prompt_id = sorted(persona.prompts, key=lambda x: x.id)[-1].id
@@ -555,19 +578,34 @@ def stream_chat_message_objects(
             reserved_message_id=reserved_message_id,
         )
 
-        if not final_msg.prompt:
-            raise RuntimeError("No Prompt found")
-
-        prompt_config = (
-            PromptConfig.from_model(
-                final_msg.prompt,
-                prompt_override=(
-                    new_msg_req.prompt_override or chat_session.prompt_override
-                ),
+        prompt_override = new_msg_req.prompt_override or chat_session.prompt_override
+        if new_msg_req.persona_override_config:
+            prompt_config = PromptConfig(
+                system_prompt=new_msg_req.persona_override_config.prompts[
+                    0
+                ].system_prompt,
+                task_prompt=new_msg_req.persona_override_config.prompts[0].task_prompt,
+                datetime_aware=new_msg_req.persona_override_config.prompts[
+                    0
+                ].datetime_aware,
+                include_citations=new_msg_req.persona_override_config.prompts[
+                    0
+                ].include_citations,
             )
-            if not persona
-            else PromptConfig.from_model(persona.prompts[0])
-        )
+        elif prompt_override:
+            if not final_msg.prompt:
+                raise ValueError(
+                    "Prompt override cannot be applied, no base prompt found."
+                )
+            prompt_config = PromptConfig.from_model(
+                final_msg.prompt,
+                prompt_override=prompt_override,
+            )
+        elif final_msg.prompt:
+            prompt_config = PromptConfig.from_model(final_msg.prompt)
+        else:
+            prompt_config = PromptConfig.from_model(persona.prompts[0])
+
         answer_style_config = AnswerStyleConfig(
             citation_config=CitationConfig(
                 all_docs_useful=selected_db_search_docs is not None
@@ -587,11 +625,13 @@ def stream_chat_message_objects(
                 answer_style_config=answer_style_config,
                 document_pruning_config=document_pruning_config,
                 retrieval_options=retrieval_options or RetrievalDetails(),
+                rerank_settings=new_msg_req.rerank_settings,
                 selected_sections=selected_sections,
                 chunks_above=new_msg_req.chunks_above,
                 chunks_below=new_msg_req.chunks_below,
                 full_doc=new_msg_req.full_doc,
                 latest_query_files=latest_query_files,
+                bypass_acl=bypass_acl,
             ),
             internet_search_tool_config=InternetSearchToolConfig(
                 answer_style_config=answer_style_config,
@@ -605,6 +645,7 @@ def stream_chat_message_objects(
                 additional_headers=custom_tool_additional_headers,
             ),
         )
+
         tools: list[Tool] = []
         for tool_list in tool_dict.values():
             tools.extend(tool_list)
@@ -736,6 +777,8 @@ def stream_chat_message_objects(
                             response=custom_tool_response.tool_result,
                             tool_name=custom_tool_response.tool_name,
                         )
+                elif packet.id == SEARCH_DOC_CONTENT_ID and include_contexts:
+                    yield cast(DanswerContexts, packet.response)
 
             elif isinstance(packet, StreamStopInfo):
                 pass
@@ -844,3 +887,30 @@ def stream_chat_message(
         )
         for obj in objects:
             yield get_json_line(obj.model_dump())
+
+
+@log_function_time()
+def gather_stream_for_slack(
+    packets: ChatPacketStream,
+) -> ChatDanswerBotResponse:
+    response = ChatDanswerBotResponse()
+
+    answer = ""
+    for packet in packets:
+        if isinstance(packet, DanswerAnswerPiece) and packet.answer_piece:
+            answer += packet.answer_piece
+        elif isinstance(packet, QADocsResponse):
+            response.docs = packet
+        elif isinstance(packet, StreamingError):
+            response.error_msg = packet.error
+        elif isinstance(packet, ChatMessageDetail):
+            response.chat_message_id = packet.message_id
+        elif isinstance(packet, LLMRelevanceFilterResponse):
+            response.llm_selected_doc_indices = packet.llm_selected_doc_indices
+        elif isinstance(packet, AllCitations):
+            response.citations = packet.citations
+
+    if answer:
+        response.answer = answer
+
+    return response

backend/danswer/configs/app_configs.py

@@ -308,6 +308,22 @@ CONFLUENCE_CONNECTOR_ATTACHMENT_CHAR_COUNT_THRESHOLD = int(
     os.environ.get("CONFLUENCE_CONNECTOR_ATTACHMENT_CHAR_COUNT_THRESHOLD", 200_000)
 )
 
+# Due to breakages in the confluence API, the timezone offset must be specified client side
+# to match the user's specified timezone.
+
+# The current state of affairs:
+# CQL queries are parsed in the user's timezone and cannot be specified in UTC
+# no API retrieves the user's timezone
+# All data is returned in UTC, so we can't derive the user's timezone from that
+
+# https://community.developer.atlassian.com/t/confluence-cloud-time-zone-get-via-rest-api/35954/16
+# https://jira.atlassian.com/browse/CONFCLOUD-69670
+
+# enter as a floating point offset from UTC in hours (-24 < val < 24)
+# this will be applied globally, so it probably makes sense to transition this to per
+# connector as some point.
+CONFLUENCE_TIMEZONE_OFFSET = float(os.environ.get("CONFLUENCE_TIMEZONE_OFFSET", 0.0))
+
 JIRA_CONNECTOR_LABELS_TO_SKIP = [
     ignored_tag
     for ignored_tag in os.environ.get("JIRA_CONNECTOR_LABELS_TO_SKIP", "").split(",")
@@ -493,10 +509,6 @@ CONTROL_PLANE_API_BASE_URL = os.environ.get(
 # JWT configuration
 JWT_ALGORITHM = "HS256"
 
-# Super Users
-SUPER_USERS = json.loads(os.environ.get("SUPER_USERS", '["pablo@danswer.ai"]'))
-SUPER_CLOUD_API_KEY = os.environ.get("SUPER_CLOUD_API_KEY", "api_key")
-
 
 #####
 # API Key Configs
@@ -510,3 +522,6 @@ API_KEY_HASH_ROUNDS = (
 
 POD_NAME = os.environ.get("POD_NAME")
 POD_NAMESPACE = os.environ.get("POD_NAMESPACE")
+
+
+DEV_MODE = os.environ.get("DEV_MODE", "").lower() == "true"

backend/danswer/configs/constants.py

@@ -31,6 +31,8 @@ DISABLED_GEN_AI_MSG = (
     "You can still use Danswer as a search engine."
 )
 
+DEFAULT_PERSONA_ID = 0
+
 # Postgres connection constants for application_name
 POSTGRES_WEB_APP_NAME = "web"
 POSTGRES_INDEXER_APP_NAME = "indexer"
@@ -259,6 +261,32 @@ class DanswerCeleryPriority(int, Enum):
     LOWEST = auto()
 
 
+class DanswerCeleryTask:
+    CHECK_FOR_CONNECTOR_DELETION = "check_for_connector_deletion_task"
+    CHECK_FOR_VESPA_SYNC_TASK = "check_for_vespa_sync_task"
+    CHECK_FOR_INDEXING = "check_for_indexing"
+    CHECK_FOR_PRUNING = "check_for_pruning"
+    CHECK_FOR_DOC_PERMISSIONS_SYNC = "check_for_doc_permissions_sync"
+    CHECK_FOR_EXTERNAL_GROUP_SYNC = "check_for_external_group_sync"
+    MONITOR_VESPA_SYNC = "monitor_vespa_sync"
+    KOMBU_MESSAGE_CLEANUP_TASK = "kombu_message_cleanup_task"
+    CONNECTOR_PERMISSION_SYNC_GENERATOR_TASK = (
+        "connector_permission_sync_generator_task"
+    )
+    UPDATE_EXTERNAL_DOCUMENT_PERMISSIONS_TASK = (
+        "update_external_document_permissions_task"
+    )
+    CONNECTOR_EXTERNAL_GROUP_SYNC_GENERATOR_TASK = (
+        "connector_external_group_sync_generator_task"
+    )
+    CONNECTOR_INDEXING_PROXY_TASK = "connector_indexing_proxy_task"
+    CONNECTOR_PRUNING_GENERATOR_TASK = "connector_pruning_generator_task"
+    DOCUMENT_BY_CC_PAIR_CLEANUP_TASK = "document_by_cc_pair_cleanup_task"
+    VESPA_METADATA_SYNC_TASK = "vespa_metadata_sync_task"
+    CHECK_TTL_MANAGEMENT_TASK = "check_ttl_management_task"
+    AUTOGENERATE_USAGE_REPORT_TASK = "autogenerate_usage_report_task"
+
+
 REDIS_SOCKET_KEEPALIVE_OPTIONS = {}
 REDIS_SOCKET_KEEPALIVE_OPTIONS[socket.TCP_KEEPINTVL] = 15
 REDIS_SOCKET_KEEPALIVE_OPTIONS[socket.TCP_KEEPCNT] = 3

backend/danswer/configs/danswerbot_configs.py

@@ -4,11 +4,8 @@ import os
 # Danswer Slack Bot Configs
 #####
 DANSWER_BOT_NUM_RETRIES = int(os.environ.get("DANSWER_BOT_NUM_RETRIES", "5"))
-DANSWER_BOT_ANSWER_GENERATION_TIMEOUT = int(
-    os.environ.get("DANSWER_BOT_ANSWER_GENERATION_TIMEOUT", "90")
-)
 # How much of the available input context can be used for thread context
-DANSWER_BOT_TARGET_CHUNK_PERCENTAGE = 512 * 2 / 3072
+MAX_THREAD_CONTEXT_PERCENTAGE = 512 * 2 / 3072
 # Number of docs to display in "Reference Documents"
 DANSWER_BOT_NUM_DOCS_TO_DISPLAY = int(
     os.environ.get("DANSWER_BOT_NUM_DOCS_TO_DISPLAY", "5")
@@ -47,17 +44,6 @@ DANSWER_BOT_DISPLAY_ERROR_MSGS = os.environ.get(
 DANSWER_BOT_RESPOND_EVERY_CHANNEL = (
     os.environ.get("DANSWER_BOT_RESPOND_EVERY_CHANNEL", "").lower() == "true"
 )
-# Add a second LLM call post Answer to verify if the Answer is valid
-# Throws out answers that don't directly or fully answer the user query
-# This is the default for all DanswerBot channels unless the channel is configured individually
-# Set/unset by "Hide Non Answers"
-ENABLE_DANSWERBOT_REFLEXION = (
-    os.environ.get("ENABLE_DANSWERBOT_REFLEXION", "").lower() == "true"
-)
-# Currently not support chain of thought, probably will add back later
-DANSWER_BOT_DISABLE_COT = True
-# if set, will default DanswerBot to use quotes and reference documents
-DANSWER_BOT_USE_QUOTES = os.environ.get("DANSWER_BOT_USE_QUOTES", "").lower() == "true"
 
 # Maximum Questions Per Minute, Default Uncapped
 DANSWER_BOT_MAX_QPM = int(os.environ.get("DANSWER_BOT_MAX_QPM") or 0) or None

backend/danswer/configs/model_configs.py

@@ -70,7 +70,9 @@ GEN_AI_NUM_RESERVED_OUTPUT_TOKENS = int(
 )
 
 # Typically, GenAI models nowadays are at least 4K tokens
-GEN_AI_MODEL_FALLBACK_MAX_TOKENS = 4096
+GEN_AI_MODEL_FALLBACK_MAX_TOKENS = int(
+    os.environ.get("GEN_AI_MODEL_FALLBACK_MAX_TOKENS") or 4096
+)
 
 # Number of tokens from chat history to include at maximum
 # 3000 should be enough context regardless of use, no need to include as much as possible

backend/danswer/connectors/README.md

@@ -11,11 +11,16 @@ Connectors come in 3 different flows:
 - Load Connector:
   - Bulk indexes documents to reflect a point in time. This type of connector generally works by either pulling all
   documents via a connector's API or loads the documents from some sort of a dump file.
-- Poll connector:
+- Poll Connector:
   - Incrementally updates documents based on a provided time range. It is used by the background job to pull the latest
   changes and additions since the last round of polling. This connector helps keep the document index up to date
   without needing to fetch/embed/index every document which would be too slow to do frequently on large sets of
   documents.
+- Slim Connector:
+  - This connector should be a lighter weight method of checking all documents in the source to see if they still exist.
+  - This connector should be identical to the Poll or Load Connector except that it only fetches the IDs of the documents, not the documents themselves.
+  - This is used by our pruning job which removes old documents from the index.
+  - The optional start and end datetimes can be ignored.
 - Event Based connectors:
   - Connectors that listen to events and update documents accordingly.
   - Currently not used by the background job, this exists for future design purposes.
@@ -26,8 +31,14 @@ Refer to [interfaces.py](https://github.com/danswer-ai/danswer/blob/main/backend
 and this first contributor created Pull Request for a new connector (Shoutout to Dan Brown):
 [Reference Pull Request](https://github.com/danswer-ai/danswer/pull/139)
 
+For implementing a Slim Connector, refer to the comments in this PR:
+[Slim Connector PR](https://github.com/danswer-ai/danswer/pull/3303/files)
+
+All new connectors should have tests added to the `backend/tests/daily/connectors` directory. Refer to the above PR for an example of adding tests for a new connector.
+
+
 #### Implementing the new Connector
-The connector must subclass one or more of LoadConnector, PollConnector, or EventConnector.
+The connector must subclass one or more of LoadConnector, PollConnector, SlimConnector, or EventConnector.
 
 The `__init__` should take arguments for configuring what documents the connector will and where it finds those
 documents. For example, if you have a wiki site, it may include the configuration for the team, topic, folder, etc. of

backend/danswer/connectors/confluence/connector.py

@@ -1,9 +1,11 @@
 from datetime import datetime
+from datetime import timedelta
 from datetime import timezone
 from typing import Any
 from urllib.parse import quote
 
 from danswer.configs.app_configs import CONFLUENCE_CONNECTOR_LABELS_TO_SKIP
+from danswer.configs.app_configs import CONFLUENCE_TIMEZONE_OFFSET
 from danswer.configs.app_configs import CONTINUE_ON_CONNECTOR_FAILURE
 from danswer.configs.app_configs import INDEX_BATCH_SIZE
 from danswer.configs.constants import DocumentSource
@@ -51,7 +53,7 @@ _RESTRICTIONS_EXPANSION_FIELDS = [
     "restrictions.read.restrictions.group",
 ]
 
-_SLIM_DOC_BATCH_SIZE = 1000
+_SLIM_DOC_BATCH_SIZE = 5000
 
 
 class ConfluenceConnector(LoadConnector, PollConnector, SlimConnector):
@@ -69,6 +71,7 @@ class ConfluenceConnector(LoadConnector, PollConnector, SlimConnector):
         # skip it. This is generally used to avoid indexing extra sensitive
         # pages.
         labels_to_skip: list[str] = CONFLUENCE_CONNECTOR_LABELS_TO_SKIP,
+        timezone_offset: float = CONFLUENCE_TIMEZONE_OFFSET,
     ) -> None:
         self.batch_size = batch_size
         self.continue_on_failure = continue_on_failure
@@ -104,6 +107,8 @@ class ConfluenceConnector(LoadConnector, PollConnector, SlimConnector):
             )
             self.cql_label_filter = f" and label not in ({comma_separated_labels})"
 
+        self.timezone: timezone = timezone(offset=timedelta(hours=timezone_offset))
+
     @property
     def confluence_client(self) -> OnyxConfluence:
         if self._confluence_client is None:
@@ -204,12 +209,14 @@ class ConfluenceConnector(LoadConnector, PollConnector, SlimConnector):
         confluence_page_ids: list[str] = []
 
         page_query = self.cql_page_query + self.cql_label_filter + self.cql_time_filter
+        logger.debug(f"page_query: {page_query}")
         # Fetch pages as Documents
         for page in self.confluence_client.paginated_cql_retrieval(
             cql=page_query,
             expand=",".join(_PAGE_EXPANSION_FIELDS),
             limit=self.batch_size,
         ):
+            logger.debug(f"_fetch_document_batches: {page['id']}")
             confluence_page_ids.append(page["id"])
             doc = self._convert_object_to_document(page)
             if doc is not None:
@@ -242,10 +249,10 @@ class ConfluenceConnector(LoadConnector, PollConnector, SlimConnector):
 
     def poll_source(self, start: float, end: float) -> GenerateDocumentsOutput:
         # Add time filters
-        formatted_start_time = datetime.fromtimestamp(start, tz=timezone.utc).strftime(
+        formatted_start_time = datetime.fromtimestamp(start, tz=self.timezone).strftime(
             "%Y-%m-%d %H:%M"
         )
-        formatted_end_time = datetime.fromtimestamp(end, tz=timezone.utc).strftime(
+        formatted_end_time = datetime.fromtimestamp(end, tz=self.timezone).strftime(
             "%Y-%m-%d %H:%M"
         )
         self.cql_time_filter = f" and lastmodified >= '{formatted_start_time}'"
@@ -301,5 +308,8 @@ class ConfluenceConnector(LoadConnector, PollConnector, SlimConnector):
                         perm_sync_data=perm_sync_data,
                     )
                 )
-            yield doc_metadata_list
-            doc_metadata_list = []
+            if len(doc_metadata_list) > _SLIM_DOC_BATCH_SIZE:
+                yield doc_metadata_list[:_SLIM_DOC_BATCH_SIZE]
+                doc_metadata_list = doc_metadata_list[_SLIM_DOC_BATCH_SIZE:]
+
+        yield doc_metadata_list

backend/danswer/connectors/confluence/onyx_confluence.py

@@ -120,7 +120,7 @@ def handle_confluence_rate_limit(confluence_call: F) -> F:
     return cast(F, wrapped_call)
 
 
-_DEFAULT_PAGINATION_LIMIT = 100
+_DEFAULT_PAGINATION_LIMIT = 1000
 
 
 class OnyxConfluence(Confluence):
@@ -134,6 +134,32 @@ class OnyxConfluence(Confluence):
         super(OnyxConfluence, self).__init__(url, *args, **kwargs)
         self._wrap_methods()
 
+    def get_current_user(self, expand: str | None = None) -> Any:
+        """
+        Implements a method that isn't in the third party client.
+
+        Get information about the current user
+        :param expand: OPTIONAL expand for get status of user.
+                Possible param is "status". Results are "Active, Deactivated"
+        :return: Returns the user details
+        """
+
+        from atlassian.errors import ApiPermissionError  # type:ignore
+
+        url = "rest/api/user/current"
+        params = {}
+        if expand:
+            params["expand"] = expand
+        try:
+            response = self.get(url, params=params)
+        except HTTPError as e:
+            if e.response.status_code == 403:
+                raise ApiPermissionError(
+                    "The calling user does not have permission", reason=e
+                )
+            raise
+        return response
+
     def _wrap_methods(self) -> None:
         """
         For each attribute that is callable (i.e., a method) and doesn't start with an underscore,
@@ -306,6 +332,13 @@ def _validate_connector_configuration(
     )
     spaces = confluence_client_with_minimal_retries.get_all_spaces(limit=1)
 
+    # uncomment the following for testing
+    # the following is an attempt to retrieve the user's timezone
+    # Unfornately, all data is returned in UTC regardless of the user's time zone
+    # even tho CQL parses incoming times based on the user's time zone
+    # space_key = spaces["results"][0]["key"]
+    # space_details = confluence_client_with_minimal_retries.cql(f"space.key={space_key}+AND+type=space")
+
     if not spaces:
         raise RuntimeError(
             f"No spaces found at {wiki_base}! "

backend/danswer/connectors/confluence/utils.py

@@ -32,7 +32,11 @@ def get_user_email_from_username__server(
             response = confluence_client.get_mobile_parameters(user_name)
             email = response.get("email")
         except Exception:
-            email = None
+            # For now, we'll just return a string that indicates failure
+            # We may want to revert to returning None in the future
+            # email = None
+            email = f"FAILED TO GET CONFLUENCE EMAIL FOR {user_name}"
+            logger.warning(f"failed to get confluence email for {user_name}")
         _USER_EMAIL_CACHE[user_name] = email
     return _USER_EMAIL_CACHE[user_name]
 

backend/danswer/connectors/slab/connector.py

@@ -12,12 +12,15 @@ from dateutil import parser
 from danswer.configs.app_configs import INDEX_BATCH_SIZE
 from danswer.configs.constants import DocumentSource
 from danswer.connectors.interfaces import GenerateDocumentsOutput
+from danswer.connectors.interfaces import GenerateSlimDocumentOutput
 from danswer.connectors.interfaces import LoadConnector
 from danswer.connectors.interfaces import PollConnector
 from danswer.connectors.interfaces import SecondsSinceUnixEpoch
+from danswer.connectors.interfaces import SlimConnector
 from danswer.connectors.models import ConnectorMissingCredentialError
 from danswer.connectors.models import Document
 from danswer.connectors.models import Section
+from danswer.connectors.models import SlimDocument
 from danswer.utils.logger import setup_logger
 
 
@@ -28,6 +31,8 @@ logger = setup_logger()
 SLAB_GRAPHQL_MAX_TRIES = 10
 SLAB_API_URL = "https://api.slab.com/v1/graphql"
 
+_SLIM_BATCH_SIZE = 1000
+
 
 def run_graphql_request(
     graphql_query: dict, bot_token: str, max_tries: int = SLAB_GRAPHQL_MAX_TRIES
@@ -158,21 +163,26 @@ def get_slab_url_from_title_id(base_url: str, title: str, page_id: str) -> str:
     return urljoin(urljoin(base_url, "posts/"), url_id)
 
 
-class SlabConnector(LoadConnector, PollConnector):
+class SlabConnector(LoadConnector, PollConnector, SlimConnector):
     def __init__(
         self,
         base_url: str,
         batch_size: int = INDEX_BATCH_SIZE,
-        slab_bot_token: str | None = None,
     ) -> None:
         self.base_url = base_url
         self.batch_size = batch_size
-        self.slab_bot_token = slab_bot_token
+        self._slab_bot_token: str | None = None
 
     def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
-        self.slab_bot_token = credentials["slab_bot_token"]
+        self._slab_bot_token = credentials["slab_bot_token"]
         return None
 
+    @property
+    def slab_bot_token(self) -> str:
+        if self._slab_bot_token is None:
+            raise ConnectorMissingCredentialError("Slab")
+        return self._slab_bot_token
+
     def _iterate_posts(
         self, time_filter: Callable[[datetime], bool] | None = None
     ) -> GenerateDocumentsOutput:
@@ -227,3 +237,21 @@ class SlabConnector(LoadConnector, PollConnector):
         yield from self._iterate_posts(
             time_filter=lambda t: start_time <= t <= end_time
         )
+
+    def retrieve_all_slim_documents(
+        self,
+        start: SecondsSinceUnixEpoch | None = None,
+        end: SecondsSinceUnixEpoch | None = None,
+    ) -> GenerateSlimDocumentOutput:
+        slim_doc_batch: list[SlimDocument] = []
+        for post_id in get_all_post_ids(self.slab_bot_token):
+            slim_doc_batch.append(
+                SlimDocument(
+                    id=post_id,
+                )
+            )
+            if len(slim_doc_batch) >= _SLIM_BATCH_SIZE:
+                yield slim_doc_batch
+                slim_doc_batch = []
+        if slim_doc_batch:
+            yield slim_doc_batch

backend/danswer/danswerbot/slack/blocks.py

@@ -16,24 +16,31 @@ from slack_sdk.models.blocks import SectionBlock
 from slack_sdk.models.blocks.basic_components import MarkdownTextObject
 from slack_sdk.models.blocks.block_elements import ImageElement
 
-from danswer.chat.models import DanswerQuote
+from danswer.chat.models import ChatDanswerBotResponse
 from danswer.configs.app_configs import DISABLE_GENERATIVE_AI
+from danswer.configs.app_configs import WEB_DOMAIN
 from danswer.configs.constants import DocumentSource
 from danswer.configs.constants import SearchFeedbackType
 from danswer.configs.danswerbot_configs import DANSWER_BOT_NUM_DOCS_TO_DISPLAY
 from danswer.context.search.models import SavedSearchDoc
+from danswer.danswerbot.slack.constants import CONTINUE_IN_WEB_UI_ACTION_ID
 from danswer.danswerbot.slack.constants import DISLIKE_BLOCK_ACTION_ID
 from danswer.danswerbot.slack.constants import FEEDBACK_DOC_BUTTON_BLOCK_ACTION_ID
 from danswer.danswerbot.slack.constants import FOLLOWUP_BUTTON_ACTION_ID
 from danswer.danswerbot.slack.constants import FOLLOWUP_BUTTON_RESOLVED_ACTION_ID
 from danswer.danswerbot.slack.constants import IMMEDIATE_RESOLVED_BUTTON_ACTION_ID
 from danswer.danswerbot.slack.constants import LIKE_BLOCK_ACTION_ID
+from danswer.danswerbot.slack.formatting import format_slack_message
 from danswer.danswerbot.slack.icons import source_to_github_img_link
+from danswer.danswerbot.slack.models import SlackMessageInfo
+from danswer.danswerbot.slack.utils import build_continue_in_web_ui_id
 from danswer.danswerbot.slack.utils import build_feedback_id
 from danswer.danswerbot.slack.utils import remove_slack_text_interactions
 from danswer.danswerbot.slack.utils import translate_vespa_highlight_to_slack
+from danswer.db.chat import get_chat_session_by_message_id
+from danswer.db.engine import get_session_with_tenant
+from danswer.db.models import ChannelConfig
 from danswer.utils.text_processing import decode_escapes
-from danswer.utils.text_processing import replace_whitespaces_w_space
 
 _MAX_BLURB_LEN = 45
 
@@ -101,12 +108,12 @@ def _split_text(text: str, limit: int = 3000) -> list[str]:
     return chunks
 
 
-def clean_markdown_link_text(text: str) -> str:
+def _clean_markdown_link_text(text: str) -> str:
     # Remove any newlines within the text
     return text.replace("\n", " ").strip()
 
 
-def build_qa_feedback_block(
+def _build_qa_feedback_block(
     message_id: int, feedback_reminder_id: str | None = None
 ) -> Block:
     return ActionsBlock(
@@ -115,7 +122,6 @@ def build_qa_feedback_block(
             ButtonElement(
                 action_id=LIKE_BLOCK_ACTION_ID,
                 text="👍 Helpful",
-                style="primary",
                 value=feedback_reminder_id,
             ),
             ButtonElement(
@@ -155,7 +161,7 @@ def get_document_feedback_blocks() -> Block:
     )
 
 
-def build_doc_feedback_block(
+def _build_doc_feedback_block(
     message_id: int,
     document_id: str,
     document_rank: int,
@@ -182,7 +188,7 @@ def get_restate_blocks(
     ]
 
 
-def build_documents_blocks(
+def _build_documents_blocks(
     documents: list[SavedSearchDoc],
     message_id: int | None,
     num_docs_to_display: int = DANSWER_BOT_NUM_DOCS_TO_DISPLAY,
@@ -223,7 +229,7 @@ def build_documents_blocks(
 
         feedback: ButtonElement | dict = {}
         if message_id is not None:
-            feedback = build_doc_feedback_block(
+            feedback = _build_doc_feedback_block(
                 message_id=message_id,
                 document_id=d.document_id,
                 document_rank=rank,
@@ -241,7 +247,7 @@ def build_documents_blocks(
     return section_blocks
 
 
-def build_sources_blocks(
+def _build_sources_blocks(
     cited_documents: list[tuple[int, SavedSearchDoc]],
     num_docs_to_display: int = DANSWER_BOT_NUM_DOCS_TO_DISPLAY,
 ) -> list[Block]:
@@ -286,7 +292,7 @@ def build_sources_blocks(
             + ([days_ago_str] if days_ago_str else [])
         )
 
-        document_title = clean_markdown_link_text(doc_sem_id)
+        document_title = _clean_markdown_link_text(doc_sem_id)
         img_link = source_to_github_img_link(d.source_type)
 
         section_blocks.append(
@@ -317,106 +323,105 @@ def build_sources_blocks(
     return section_blocks
 
 
-def build_quotes_block(
-    quotes: list[DanswerQuote],
+def _priority_ordered_documents_blocks(
+    answer: ChatDanswerBotResponse,
 ) -> list[Block]:
-    quote_lines: list[str] = []
-    doc_to_quotes: dict[str, list[str]] = {}
-    doc_to_link: dict[str, str] = {}
-    doc_to_sem_id: dict[str, str] = {}
-    for q in quotes:
-        quote = q.quote
-        doc_id = q.document_id
-        doc_link = q.link
-        doc_name = q.semantic_identifier
-        if doc_link and doc_name and doc_id and quote:
-            if doc_id not in doc_to_quotes:
-                doc_to_quotes[doc_id] = [quote]
-                doc_to_link[doc_id] = doc_link
-                doc_to_sem_id[doc_id] = (
-                    doc_name
-                    if q.source_type != DocumentSource.SLACK.value
-                    else "#" + doc_name
-                )
-            else:
-                doc_to_quotes[doc_id].append(quote)
-
-    for doc_id, quote_strs in doc_to_quotes.items():
-        quotes_str_clean = [
-            replace_whitespaces_w_space(q_str).strip() for q_str in quote_strs
-        ]
-        longest_quotes = sorted(quotes_str_clean, key=len, reverse=True)[:5]
-        single_quote_str = "\n".join([f"```{q_str}```" for q_str in longest_quotes])
-        link = doc_to_link[doc_id]
-        sem_id = doc_to_sem_id[doc_id]
-        quote_lines.append(
-            f"<{link}|{sem_id}>:\n{remove_slack_text_interactions(single_quote_str)}"
-        )
-
-    if not doc_to_quotes:
+    docs_response = answer.docs if answer.docs else None
+    top_docs = docs_response.top_documents if docs_response else []
+    llm_doc_inds = answer.llm_selected_doc_indices or []
+    llm_docs = [top_docs[i] for i in llm_doc_inds]
+    remaining_docs = [
+        doc for idx, doc in enumerate(top_docs) if idx not in llm_doc_inds
+    ]
+    priority_ordered_docs = llm_docs + remaining_docs
+    if not priority_ordered_docs:
         return []
 
-    return [SectionBlock(text="*Relevant Snippets*\n" + "\n".join(quote_lines))]
+    document_blocks = _build_documents_blocks(
+        documents=priority_ordered_docs,
+        message_id=answer.chat_message_id,
+    )
+    if document_blocks:
+        document_blocks = [DividerBlock()] + document_blocks
+    return document_blocks
 
 
-def build_qa_response_blocks(
-    message_id: int | None,
-    answer: str | None,
-    quotes: list[DanswerQuote] | None,
-    source_filters: list[DocumentSource] | None,
-    time_cutoff: datetime | None,
-    favor_recent: bool,
-    skip_quotes: bool = False,
-    process_message_for_citations: bool = False,
-    skip_ai_feedback: bool = False,
-    feedback_reminder_id: str | None = None,
+def _build_citations_blocks(
+    answer: ChatDanswerBotResponse,
 ) -> list[Block]:
+    docs_response = answer.docs if answer.docs else None
+    top_docs = docs_response.top_documents if docs_response else []
+    citations = answer.citations or []
+    cited_docs = []
+    for citation in citations:
+        matching_doc = next(
+            (d for d in top_docs if d.document_id == citation.document_id),
+            None,
+        )
+        if matching_doc:
+            cited_docs.append((citation.citation_num, matching_doc))
+
+    cited_docs.sort()
+    citations_block = _build_sources_blocks(cited_documents=cited_docs)
+    return citations_block
+
+
+def _build_qa_response_blocks(
+    answer: ChatDanswerBotResponse,
+    process_message_for_citations: bool = False,
+) -> list[Block]:
+    retrieval_info = answer.docs
+    if not retrieval_info:
+        # This should not happen, even with no docs retrieved, there is still info returned
+        raise RuntimeError("Failed to retrieve docs, cannot answer question.")
+
+    formatted_answer = format_slack_message(answer.answer) if answer.answer else None
+
     if DISABLE_GENERATIVE_AI:
         return []
 
-    quotes_blocks: list[Block] = []
-
     filter_block: Block | None = None
-    if time_cutoff or favor_recent or source_filters:
+    if (
+        retrieval_info.applied_time_cutoff
+        or retrieval_info.recency_bias_multiplier > 1
+        or retrieval_info.applied_source_filters
+    ):
         filter_text = "Filters: "
-        if source_filters:
-            sources_str = ", ".join([s.value for s in source_filters])
+        if retrieval_info.applied_source_filters:
+            sources_str = ", ".join(
+                [s.value for s in retrieval_info.applied_source_filters]
+            )
             filter_text += f"`Sources in [{sources_str}]`"
-            if time_cutoff or favor_recent:
+            if (
+                retrieval_info.applied_time_cutoff
+                or retrieval_info.recency_bias_multiplier > 1
+            ):
                 filter_text += " and "
-        if time_cutoff is not None:
-            time_str = time_cutoff.strftime("%b %d, %Y")
+        if retrieval_info.applied_time_cutoff is not None:
+            time_str = retrieval_info.applied_time_cutoff.strftime("%b %d, %Y")
             filter_text += f"`Docs Updated >= {time_str}` "
-        if favor_recent:
-            if time_cutoff is not None:
+        if retrieval_info.recency_bias_multiplier > 1:
+            if retrieval_info.applied_time_cutoff is not None:
                 filter_text += "+ "
             filter_text += "`Prioritize Recently Updated Docs`"
 
         filter_block = SectionBlock(text=f"_{filter_text}_")
 
-    if not answer:
+    if not formatted_answer:
         answer_blocks = [
             SectionBlock(
                 text="Sorry, I was unable to find an answer, but I did find some potentially relevant docs 🤓"
             )
         ]
     else:
-        answer_processed = decode_escapes(remove_slack_text_interactions(answer))
+        answer_processed = decode_escapes(
+            remove_slack_text_interactions(formatted_answer)
+        )
         if process_message_for_citations:
             answer_processed = _process_citations_for_slack(answer_processed)
         answer_blocks = [
             SectionBlock(text=text) for text in _split_text(answer_processed)
         ]
-        if quotes:
-            quotes_blocks = build_quotes_block(quotes)
-
-        # if no quotes OR `build_quotes_block()` did not give back any blocks
-        if not quotes_blocks:
-            quotes_blocks = [
-                SectionBlock(
-                    text="*Warning*: no sources were quoted for this answer, so it may be unreliable 😔"
-                )
-            ]
 
     response_blocks: list[Block] = []
 
@@ -425,20 +430,34 @@ def build_qa_response_blocks(
 
     response_blocks.extend(answer_blocks)
 
-    if message_id is not None and not skip_ai_feedback:
-        response_blocks.append(
-            build_qa_feedback_block(
-                message_id=message_id, feedback_reminder_id=feedback_reminder_id
-            )
-        )
-
-    if not skip_quotes:
-        response_blocks.extend(quotes_blocks)
-
     return response_blocks
 
 
-def build_follow_up_block(message_id: int | None) -> ActionsBlock:
+def _build_continue_in_web_ui_block(
+    tenant_id: str | None,
+    message_id: int | None,
+) -> Block:
+    if message_id is None:
+        raise ValueError("No message id provided to build continue in web ui block")
+    with get_session_with_tenant(tenant_id) as db_session:
+        chat_session = get_chat_session_by_message_id(
+            db_session=db_session,
+            message_id=message_id,
+        )
+        return ActionsBlock(
+            block_id=build_continue_in_web_ui_id(message_id),
+            elements=[
+                ButtonElement(
+                    action_id=CONTINUE_IN_WEB_UI_ACTION_ID,
+                    text="Continue Chat in Danswer!",
+                    style="primary",
+                    url=f"{WEB_DOMAIN}/chat?slackChatId={chat_session.id}",
+                ),
+            ],
+        )
+
+
+def _build_follow_up_block(message_id: int | None) -> ActionsBlock:
     return ActionsBlock(
         block_id=build_feedback_id(message_id) if message_id is not None else None,
         elements=[
@@ -483,3 +502,75 @@ def build_follow_up_resolved_blocks(
         ]
     )
     return [text_block, button_block]
+
+
+def build_slack_response_blocks(
+    answer: ChatDanswerBotResponse,
+    tenant_id: str | None,
+    message_info: SlackMessageInfo,
+    channel_conf: ChannelConfig | None,
+    use_citations: bool,
+    feedback_reminder_id: str | None,
+    skip_ai_feedback: bool = False,
+) -> list[Block]:
+    """
+    This function is a top level function that builds all the blocks for the Slack response.
+    It also handles combining all the blocks together.
+    """
+    # If called with the DanswerBot slash command, the question is lost so we have to reshow it
+    restate_question_block = get_restate_blocks(
+        message_info.thread_messages[-1].message, message_info.is_bot_msg
+    )
+
+    answer_blocks = _build_qa_response_blocks(
+        answer=answer,
+        process_message_for_citations=use_citations,
+    )
+
+    web_follow_up_block = []
+    if channel_conf and channel_conf.get("show_continue_in_web_ui"):
+        web_follow_up_block.append(
+            _build_continue_in_web_ui_block(
+                tenant_id=tenant_id,
+                message_id=answer.chat_message_id,
+            )
+        )
+
+    follow_up_block = []
+    if channel_conf and channel_conf.get("follow_up_tags") is not None:
+        follow_up_block.append(
+            _build_follow_up_block(message_id=answer.chat_message_id)
+        )
+
+    ai_feedback_block = []
+    if answer.chat_message_id is not None and not skip_ai_feedback:
+        ai_feedback_block.append(
+            _build_qa_feedback_block(
+                message_id=answer.chat_message_id,
+                feedback_reminder_id=feedback_reminder_id,
+            )
+        )
+
+    citations_blocks = []
+    document_blocks = []
+    if use_citations and answer.citations:
+        citations_blocks = _build_citations_blocks(answer)
+    else:
+        document_blocks = _priority_ordered_documents_blocks(answer)
+
+    citations_divider = [DividerBlock()] if citations_blocks else []
+    buttons_divider = [DividerBlock()] if web_follow_up_block or follow_up_block else []
+
+    all_blocks = (
+        restate_question_block
+        + answer_blocks
+        + ai_feedback_block
+        + citations_divider
+        + citations_blocks
+        + document_blocks
+        + buttons_divider
+        + web_follow_up_block
+        + follow_up_block
+    )
+
+    return all_blocks

backend/danswer/danswerbot/slack/constants.py

@@ -2,6 +2,7 @@ from enum import Enum
 
 LIKE_BLOCK_ACTION_ID = "feedback-like"
 DISLIKE_BLOCK_ACTION_ID = "feedback-dislike"
+CONTINUE_IN_WEB_UI_ACTION_ID = "continue-in-web-ui"
 FEEDBACK_DOC_BUTTON_BLOCK_ACTION_ID = "feedback-doc-button"
 IMMEDIATE_RESOLVED_BUTTON_ACTION_ID = "immediate-resolved-button"
 FOLLOWUP_BUTTON_ACTION_ID = "followup-button"

backend/danswer/danswerbot/slack/handlers/handle_buttons.py

@@ -28,7 +28,7 @@ from danswer.danswerbot.slack.models import SlackMessageInfo
 from danswer.danswerbot.slack.utils import build_feedback_id
 from danswer.danswerbot.slack.utils import decompose_action_id
 from danswer.danswerbot.slack.utils import fetch_group_ids_from_names
-from danswer.danswerbot.slack.utils import fetch_user_ids_from_emails
+from danswer.danswerbot.slack.utils import fetch_slack_user_ids_from_emails
 from danswer.danswerbot.slack.utils import get_channel_name_from_id
 from danswer.danswerbot.slack.utils import get_feedback_visibility
 from danswer.danswerbot.slack.utils import read_slack_thread
@@ -267,7 +267,7 @@ def handle_followup_button(
             tag_names = slack_channel_config.channel_config.get("follow_up_tags")
             remaining = None
             if tag_names:
-                tag_ids, remaining = fetch_user_ids_from_emails(
+                tag_ids, remaining = fetch_slack_user_ids_from_emails(
                     tag_names, client.web_client
                 )
             if remaining:

backend/danswer/danswerbot/slack/handlers/handle_message.py

@@ -13,7 +13,7 @@ from danswer.danswerbot.slack.handlers.handle_standard_answers import (
     handle_standard_answers,
 )
 from danswer.danswerbot.slack.models import SlackMessageInfo
-from danswer.danswerbot.slack.utils import fetch_user_ids_from_emails
+from danswer.danswerbot.slack.utils import fetch_slack_user_ids_from_emails
 from danswer.danswerbot.slack.utils import fetch_user_ids_from_groups
 from danswer.danswerbot.slack.utils import respond_in_thread
 from danswer.danswerbot.slack.utils import slack_usage_report
@@ -184,7 +184,7 @@ def handle_message(
     send_to: list[str] | None = None
     missing_users: list[str] | None = None
     if respond_member_group_list:
-        send_to, missing_ids = fetch_user_ids_from_emails(
+        send_to, missing_ids = fetch_slack_user_ids_from_emails(
             respond_member_group_list, client
         )
 

backend/danswer/danswerbot/slack/handlers/handle_regular_answer.py

@@ -1,60 +1,43 @@
 import functools
 from collections.abc import Callable
 from typing import Any
-from typing import cast
 from typing import Optional
 from typing import TypeVar
 
 from retry import retry
 from slack_sdk import WebClient
-from slack_sdk.models.blocks import DividerBlock
 from slack_sdk.models.blocks import SectionBlock
 
+from danswer.chat.chat_utils import prepare_chat_message_request
+from danswer.chat.models import ChatDanswerBotResponse
+from danswer.chat.process_message import gather_stream_for_slack
+from danswer.chat.process_message import stream_chat_message_objects
 from danswer.configs.app_configs import DISABLE_GENERATIVE_AI
-from danswer.configs.danswerbot_configs import DANSWER_BOT_ANSWER_GENERATION_TIMEOUT
-from danswer.configs.danswerbot_configs import DANSWER_BOT_DISABLE_COT
+from danswer.configs.constants import DEFAULT_PERSONA_ID
 from danswer.configs.danswerbot_configs import DANSWER_BOT_DISABLE_DOCS_ONLY_ANSWER
 from danswer.configs.danswerbot_configs import DANSWER_BOT_DISPLAY_ERROR_MSGS
 from danswer.configs.danswerbot_configs import DANSWER_BOT_NUM_RETRIES
-from danswer.configs.danswerbot_configs import DANSWER_BOT_TARGET_CHUNK_PERCENTAGE
-from danswer.configs.danswerbot_configs import DANSWER_BOT_USE_QUOTES
 from danswer.configs.danswerbot_configs import DANSWER_FOLLOWUP_EMOJI
 from danswer.configs.danswerbot_configs import DANSWER_REACT_EMOJI
-from danswer.configs.danswerbot_configs import ENABLE_DANSWERBOT_REFLEXION
+from danswer.configs.danswerbot_configs import MAX_THREAD_CONTEXT_PERCENTAGE
 from danswer.context.search.enums import OptionalSearchSetting
 from danswer.context.search.models import BaseFilters
-from danswer.context.search.models import RerankingDetails
 from danswer.context.search.models import RetrievalDetails
-from danswer.danswerbot.slack.blocks import build_documents_blocks
-from danswer.danswerbot.slack.blocks import build_follow_up_block
-from danswer.danswerbot.slack.blocks import build_qa_response_blocks
-from danswer.danswerbot.slack.blocks import build_sources_blocks
-from danswer.danswerbot.slack.blocks import get_restate_blocks
-from danswer.danswerbot.slack.formatting import format_slack_message
+from danswer.danswerbot.slack.blocks import build_slack_response_blocks
 from danswer.danswerbot.slack.handlers.utils import send_team_member_message
+from danswer.danswerbot.slack.handlers.utils import slackify_message_thread
 from danswer.danswerbot.slack.models import SlackMessageInfo
 from danswer.danswerbot.slack.utils import respond_in_thread
 from danswer.danswerbot.slack.utils import SlackRateLimiter
 from danswer.danswerbot.slack.utils import update_emote_react
 from danswer.db.engine import get_session_with_tenant
-from danswer.db.models import Persona
-from danswer.db.models import SlackBotResponseType
 from danswer.db.models import SlackChannelConfig
-from danswer.db.persona import fetch_persona_by_id
-from danswer.db.search_settings import get_current_search_settings
+from danswer.db.models import User
+from danswer.db.persona import get_persona_by_id
 from danswer.db.users import get_user_by_email
-from danswer.llm.answering.prompts.citations_prompt import (
-    compute_max_document_tokens_for_persona,
-)
-from danswer.llm.factory import get_llms_for_persona
-from danswer.llm.utils import check_number_of_tokens
-from danswer.llm.utils import get_max_input_tokens
-from danswer.one_shot_answer.answer_question import get_search_answer
-from danswer.one_shot_answer.models import DirectQARequest
-from danswer.one_shot_answer.models import OneShotQAResponse
+from danswer.server.query_and_chat.models import CreateChatMessageRequest
 from danswer.utils.logger import DanswerLoggingAdapter
 
-
 srl = SlackRateLimiter()
 
 RT = TypeVar("RT")  # return type
@@ -89,16 +72,14 @@ def handle_regular_answer(
     feedback_reminder_id: str | None,
     tenant_id: str | None,
     num_retries: int = DANSWER_BOT_NUM_RETRIES,
-    answer_generation_timeout: int = DANSWER_BOT_ANSWER_GENERATION_TIMEOUT,
-    thread_context_percent: float = DANSWER_BOT_TARGET_CHUNK_PERCENTAGE,
+    thread_context_percent: float = MAX_THREAD_CONTEXT_PERCENTAGE,
     should_respond_with_error_msgs: bool = DANSWER_BOT_DISPLAY_ERROR_MSGS,
     disable_docs_only_answer: bool = DANSWER_BOT_DISABLE_DOCS_ONLY_ANSWER,
-    disable_cot: bool = DANSWER_BOT_DISABLE_COT,
-    reflexion: bool = ENABLE_DANSWERBOT_REFLEXION,
 ) -> bool:
     channel_conf = slack_channel_config.channel_config if slack_channel_config else None
 
     messages = message_info.thread_messages
+
     message_ts_to_respond_to = message_info.msg_to_respond
     is_bot_msg = message_info.is_bot_msg
     user = None
@@ -108,9 +89,18 @@ def handle_regular_answer(
                 user = get_user_by_email(message_info.email, db_session)
 
     document_set_names: list[str] | None = None
-    persona = slack_channel_config.persona if slack_channel_config else None
     prompt = None
-    if persona:
+    # If no persona is specified, use the default search based persona
+    # This way slack flow always has a persona
+    persona = slack_channel_config.persona if slack_channel_config else None
+    if not persona:
+        with get_session_with_tenant(tenant_id) as db_session:
+            persona = get_persona_by_id(DEFAULT_PERSONA_ID, user, db_session)
+            document_set_names = [
+                document_set.name for document_set in persona.document_sets
+            ]
+            prompt = persona.prompts[0] if persona.prompts else None
+    else:
         document_set_names = [
             document_set.name for document_set in persona.document_sets
         ]
@@ -118,6 +108,26 @@ def handle_regular_answer(
 
     should_respond_even_with_no_docs = persona.num_chunks == 0 if persona else False
 
+    # TODO: Add in support for Slack to truncate messages based on max LLM context
+    # llm, _ = get_llms_for_persona(persona)
+
+    # llm_tokenizer = get_tokenizer(
+    #     model_name=llm.config.model_name,
+    #     provider_type=llm.config.model_provider,
+    # )
+
+    # # In cases of threads, split the available tokens between docs and thread context
+    # input_tokens = get_max_input_tokens(
+    #     model_name=llm.config.model_name,
+    #     model_provider=llm.config.model_provider,
+    # )
+    # max_history_tokens = int(input_tokens * thread_context_percent)
+    # combined_message = combine_message_thread(
+    #     messages, max_tokens=max_history_tokens, llm_tokenizer=llm_tokenizer
+    # )
+
+    combined_message = slackify_message_thread(messages)
+
     bypass_acl = False
     if (
         slack_channel_config
@@ -128,13 +138,6 @@ def handle_regular_answer(
         # with non-public document sets
         bypass_acl = True
 
-    # figure out if we want to use citations or quotes
-    use_citations = (
-        not DANSWER_BOT_USE_QUOTES
-        if slack_channel_config is None
-        else slack_channel_config.response_type == SlackBotResponseType.CITATIONS
-    )
-
     if not message_ts_to_respond_to and not is_bot_msg:
         # if the message is not "/danswer" command, then it should have a message ts to respond to
         raise RuntimeError(
@@ -147,75 +150,23 @@ def handle_regular_answer(
         backoff=2,
     )
     @rate_limits(client=client, channel=channel, thread_ts=message_ts_to_respond_to)
-    def _get_answer(new_message_request: DirectQARequest) -> OneShotQAResponse | None:
-        max_document_tokens: int | None = None
-        max_history_tokens: int | None = None
-
+    def _get_slack_answer(
+        new_message_request: CreateChatMessageRequest, danswer_user: User | None
+    ) -> ChatDanswerBotResponse:
         with get_session_with_tenant(tenant_id) as db_session:
-            if len(new_message_request.messages) > 1:
-                if new_message_request.persona_config:
-                    raise RuntimeError("Slack bot does not support persona config")
-                elif new_message_request.persona_id is not None:
-                    persona = cast(
-                        Persona,
-                        fetch_persona_by_id(
-                            db_session,
-                            new_message_request.persona_id,
-                            user=None,
-                            get_editable=False,
-                        ),
-                    )
-                else:
-                    raise RuntimeError(
-                        "No persona id provided, this should never happen."
-                    )
-
-                llm, _ = get_llms_for_persona(persona)
-
-                # In cases of threads, split the available tokens between docs and thread context
-                input_tokens = get_max_input_tokens(
-                    model_name=llm.config.model_name,
-                    model_provider=llm.config.model_provider,
-                )
-                max_history_tokens = int(input_tokens * thread_context_percent)
-
-                remaining_tokens = input_tokens - max_history_tokens
-
-                query_text = new_message_request.messages[0].message
-                if persona:
-                    max_document_tokens = compute_max_document_tokens_for_persona(
-                        persona=persona,
-                        actual_user_input=query_text,
-                        max_llm_token_override=remaining_tokens,
-                    )
-                else:
-                    max_document_tokens = (
-                        remaining_tokens
-                        - 512  # Needs to be more than any of the QA prompts
-                        - check_number_of_tokens(query_text)
-                    )
-
-            if DISABLE_GENERATIVE_AI:
-                return None
-
-            # This also handles creating the query event in postgres
-            answer = get_search_answer(
-                query_req=new_message_request,
-                user=user,
-                max_document_tokens=max_document_tokens,
-                max_history_tokens=max_history_tokens,
+            packets = stream_chat_message_objects(
+                new_msg_req=new_message_request,
+                user=danswer_user,
                 db_session=db_session,
-                answer_generation_timeout=answer_generation_timeout,
-                enable_reflexion=reflexion,
                 bypass_acl=bypass_acl,
-                use_citations=use_citations,
-                danswerbot_flow=True,
             )
 
-            if not answer.error_msg:
-                return answer
-            else:
-                raise RuntimeError(answer.error_msg)
+            answer = gather_stream_for_slack(packets)
+
+        if answer.error_msg:
+            raise RuntimeError(answer.error_msg)
+
+        return answer
 
     try:
         # By leaving time_cutoff and favor_recent as None, and setting enable_auto_detect_filters
@@ -245,26 +196,24 @@ def handle_regular_answer(
             enable_auto_detect_filters=auto_detect_filters,
         )
 
-        # Always apply reranking settings if it exists, this is the non-streaming flow
         with get_session_with_tenant(tenant_id) as db_session:
-            saved_search_settings = get_current_search_settings(db_session)
-
-        # This includes throwing out answer via reflexion
-        answer = _get_answer(
-            DirectQARequest(
-                messages=messages,
-                multilingual_query_expansion=saved_search_settings.multilingual_expansion
-                if saved_search_settings
-                else None,
-                prompt_id=prompt.id if prompt else None,
-                persona_id=persona.id if persona is not None else 0,
-                retrieval_options=retrieval_details,
-                chain_of_thought=not disable_cot,
-                rerank_settings=RerankingDetails.from_db_model(saved_search_settings)
-                if saved_search_settings
-                else None,
+            answer_request = prepare_chat_message_request(
+                message_text=combined_message,
+                user=user,
+                persona_id=persona.id,
+                # This is not used in the Slack flow, only in the answer API
+                persona_override_config=None,
+                prompt=prompt,
+                message_ts_to_respond_to=message_ts_to_respond_to,
+                retrieval_details=retrieval_details,
+                rerank_settings=None,  # Rerank customization supported in Slack flow
+                db_session=db_session,
             )
+
+        answer = _get_slack_answer(
+            new_message_request=answer_request, danswer_user=user
         )
+
     except Exception as e:
         logger.exception(
             f"Unable to process message - did not successfully answer "
@@ -365,7 +314,7 @@ def handle_regular_answer(
     top_docs = retrieval_info.top_documents
     if not top_docs and not should_respond_even_with_no_docs:
         logger.error(
-            f"Unable to answer question: '{answer.rephrase}' - no documents found"
+            f"Unable to answer question: '{combined_message}' - no documents found"
         )
         # Optionally, respond in thread with the error message
         # Used primarily for debugging purposes
@@ -386,18 +335,18 @@ def handle_regular_answer(
         )
         return True
 
-    only_respond_with_citations_or_quotes = (
+    only_respond_if_citations = (
         channel_conf
         and "well_answered_postfilter" in channel_conf.get("answer_filters", [])
     )
-    has_citations_or_quotes = bool(answer.citations or answer.quotes)
+
     if (
-        only_respond_with_citations_or_quotes
-        and not has_citations_or_quotes
+        only_respond_if_citations
+        and not answer.citations
         and not message_info.bypass_filters
     ):
         logger.error(
-            f"Unable to find citations or quotes to answer: '{answer.rephrase}' - not answering!"
+            f"Unable to find citations to answer: '{answer.answer}' - not answering!"
         )
         # Optionally, respond in thread with the error message
         # Used primarily for debugging purposes
@@ -411,62 +360,15 @@ def handle_regular_answer(
             )
         return True
 
-    # If called with the DanswerBot slash command, the question is lost so we have to reshow it
-    restate_question_block = get_restate_blocks(messages[-1].message, is_bot_msg)
-    formatted_answer = format_slack_message(answer.answer) if answer.answer else None
-
-    answer_blocks = build_qa_response_blocks(
-        message_id=answer.chat_message_id,
-        answer=formatted_answer,
-        quotes=answer.quotes.quotes if answer.quotes else None,
-        source_filters=retrieval_info.applied_source_filters,
-        time_cutoff=retrieval_info.applied_time_cutoff,
-        favor_recent=retrieval_info.recency_bias_multiplier > 1,
-        # currently Personas don't support quotes
-        # if citations are enabled, also don't use quotes
-        skip_quotes=persona is not None or use_citations,
-        process_message_for_citations=use_citations,
+    all_blocks = build_slack_response_blocks(
+        tenant_id=tenant_id,
+        message_info=message_info,
+        answer=answer,
+        channel_conf=channel_conf,
+        use_citations=True,  # No longer supporting quotes
         feedback_reminder_id=feedback_reminder_id,
     )
 
-    # Get the chunks fed to the LLM only, then fill with other docs
-    llm_doc_inds = answer.llm_selected_doc_indices or []
-    llm_docs = [top_docs[i] for i in llm_doc_inds]
-    remaining_docs = [
-        doc for idx, doc in enumerate(top_docs) if idx not in llm_doc_inds
-    ]
-    priority_ordered_docs = llm_docs + remaining_docs
-
-    document_blocks = []
-    citations_block = []
-    # if citations are enabled, only show cited documents
-    if use_citations:
-        citations = answer.citations or []
-        cited_docs = []
-        for citation in citations:
-            matching_doc = next(
-                (d for d in top_docs if d.document_id == citation.document_id),
-                None,
-            )
-            if matching_doc:
-                cited_docs.append((citation.citation_num, matching_doc))
-
-        cited_docs.sort()
-        citations_block = build_sources_blocks(cited_documents=cited_docs)
-    elif priority_ordered_docs:
-        document_blocks = build_documents_blocks(
-            documents=priority_ordered_docs,
-            message_id=answer.chat_message_id,
-        )
-        document_blocks = [DividerBlock()] + document_blocks
-
-    all_blocks = (
-        restate_question_block + answer_blocks + citations_block + document_blocks
-    )
-
-    if channel_conf and channel_conf.get("follow_up_tags") is not None:
-        all_blocks.append(build_follow_up_block(message_id=answer.chat_message_id))
-
     try:
         respond_in_thread(
             client=client,

backend/danswer/danswerbot/slack/handlers/utils.py

@@ -1,8 +1,33 @@
 from slack_sdk import WebClient
 
+from danswer.chat.models import ThreadMessage
+from danswer.configs.constants import MessageType
 from danswer.danswerbot.slack.utils import respond_in_thread
 
 
+def slackify_message_thread(messages: list[ThreadMessage]) -> str:
+    # Note: this does not handle extremely long threads, every message will be included
+    # with weaker LLMs, this could cause issues with exceeeding the token limit
+    if not messages:
+        return ""
+
+    message_strs: list[str] = []
+    for message in messages:
+        if message.role == MessageType.USER:
+            message_text = (
+                f"{message.sender or 'Unknown User'} said in Slack:\n{message.message}"
+            )
+        elif message.role == MessageType.ASSISTANT:
+            message_text = f"AI said in Slack:\n{message.message}"
+        else:
+            message_text = (
+                f"{message.role.value.upper()} said in Slack:\n{message.message}"
+            )
+        message_strs.append(message_text)
+
+    return "\n\n".join(message_strs)
+
+
 def send_team_member_message(
     client: WebClient,
     channel: str,

backend/danswer/danswerbot/slack/listener.py

@@ -19,6 +19,8 @@ from slack_sdk.socket_mode.request import SocketModeRequest
 from slack_sdk.socket_mode.response import SocketModeResponse
 from sqlalchemy.orm import Session
 
+from danswer.chat.models import ThreadMessage
+from danswer.configs.app_configs import DEV_MODE
 from danswer.configs.app_configs import POD_NAME
 from danswer.configs.app_configs import POD_NAMESPACE
 from danswer.configs.constants import DanswerRedisLocks
@@ -74,7 +76,6 @@ from danswer.db.slack_bot import fetch_slack_bots
 from danswer.key_value_store.interface import KvKeyNotFoundError
 from danswer.natural_language_processing.search_nlp_models import EmbeddingModel
 from danswer.natural_language_processing.search_nlp_models import warm_up_bi_encoder
-from danswer.one_shot_answer.models import ThreadMessage
 from danswer.redis.redis_pool import get_redis_client
 from danswer.server.manage.models import SlackBotTokens
 from danswer.utils.logger import setup_logger
@@ -250,7 +251,7 @@ class SlackbotHandler:
                 nx=True,
                 ex=TENANT_LOCK_EXPIRATION,
             )
-            if not acquired:
+            if not acquired and not DEV_MODE:
                 logger.debug(f"Another pod holds the lock for tenant {tenant_id}")
                 continue
 

backend/danswer/danswerbot/slack/models.py

@@ -1,6 +1,6 @@
 from pydantic import BaseModel
 
-from danswer.one_shot_answer.models import ThreadMessage
+from danswer.chat.models import ThreadMessage
 
 
 class SlackMessageInfo(BaseModel):

backend/danswer/danswerbot/slack/utils.py

@@ -3,9 +3,9 @@ import random
 import re
 import string
 import time
+import uuid
 from typing import Any
 from typing import cast
-from typing import Optional
 
 from retry import retry
 from slack_sdk import WebClient
@@ -30,13 +30,13 @@ from danswer.configs.danswerbot_configs import (
 from danswer.connectors.slack.utils import make_slack_api_rate_limited
 from danswer.connectors.slack.utils import SlackTextCleaner
 from danswer.danswerbot.slack.constants import FeedbackVisibility
+from danswer.danswerbot.slack.models import ThreadMessage
 from danswer.db.engine import get_session_with_tenant
 from danswer.db.users import get_user_by_email
 from danswer.llm.exceptions import GenAIDisabledException
 from danswer.llm.factory import get_default_llms
 from danswer.llm.utils import dict_based_prompt_to_langchain_prompt
 from danswer.llm.utils import message_to_string
-from danswer.one_shot_answer.models import ThreadMessage
 from danswer.prompts.miscellaneous_prompts import SLACK_LANGUAGE_REPHRASE_PROMPT
 from danswer.utils.logger import setup_logger
 from danswer.utils.telemetry import optional_telemetry
@@ -216,6 +216,13 @@ def build_feedback_id(
     return unique_prefix + ID_SEPARATOR + feedback_id
 
 
+def build_continue_in_web_ui_id(
+    message_id: int,
+) -> str:
+    unique_prefix = str(uuid.uuid4())[:10]
+    return unique_prefix + ID_SEPARATOR + str(message_id)
+
+
 def decompose_action_id(feedback_id: str) -> tuple[int, str | None, int | None]:
     """Decompose into query_id, document_id, document_rank, see above function"""
     try:
@@ -313,7 +320,7 @@ def get_channel_name_from_id(
         raise e
 
 
-def fetch_user_ids_from_emails(
+def fetch_slack_user_ids_from_emails(
     user_emails: list[str], client: WebClient
 ) -> tuple[list[str], list[str]]:
     user_ids: list[str] = []
@@ -522,7 +529,7 @@ class SlackRateLimiter:
             self.last_reset_time = time.time()
 
     def notify(
-        self, client: WebClient, channel: str, position: int, thread_ts: Optional[str]
+        self, client: WebClient, channel: str, position: int, thread_ts: str | None
     ) -> None:
         respond_in_thread(
             client=client,

backend/danswer/db/chat.py

@@ -3,6 +3,7 @@ from datetime import datetime
 from datetime import timedelta
 from uuid import UUID
 
+from fastapi import HTTPException
 from sqlalchemy import delete
 from sqlalchemy import desc
 from sqlalchemy import func
@@ -30,6 +31,7 @@ from danswer.db.models import SearchDoc
 from danswer.db.models import SearchDoc as DBSearchDoc
 from danswer.db.models import ToolCall
 from danswer.db.models import User
+from danswer.db.persona import get_best_persona_id_for_user
 from danswer.db.pg_file_store import delete_lobj_by_name
 from danswer.file_store.models import FileDescriptor
 from danswer.llm.override_models import LLMOverride
@@ -143,16 +145,10 @@ def get_chat_sessions_by_user(
     user_id: UUID | None,
     deleted: bool | None,
     db_session: Session,
-    only_one_shot: bool = False,
     limit: int = 50,
 ) -> list[ChatSession]:
     stmt = select(ChatSession).where(ChatSession.user_id == user_id)
 
-    if only_one_shot:
-        stmt = stmt.where(ChatSession.one_shot.is_(True))
-    else:
-        stmt = stmt.where(ChatSession.one_shot.is_(False))
-
     stmt = stmt.order_by(desc(ChatSession.time_created))
 
     if deleted is not None:
@@ -224,12 +220,11 @@ def delete_messages_and_files_from_chat_session(
 
 def create_chat_session(
     db_session: Session,
-    description: str,
+    description: str | None,
     user_id: UUID | None,
     persona_id: int | None,  # Can be none if temporary persona is used
     llm_override: LLMOverride | None = None,
     prompt_override: PromptOverride | None = None,
-    one_shot: bool = False,
     danswerbot_flow: bool = False,
     slack_thread_id: str | None = None,
 ) -> ChatSession:
@@ -239,7 +234,6 @@ def create_chat_session(
         description=description,
         llm_override=llm_override,
         prompt_override=prompt_override,
-        one_shot=one_shot,
         danswerbot_flow=danswerbot_flow,
         slack_thread_id=slack_thread_id,
     )
@@ -250,6 +244,48 @@ def create_chat_session(
     return chat_session
 
 
+def duplicate_chat_session_for_user_from_slack(
+    db_session: Session,
+    user: User | None,
+    chat_session_id: UUID,
+) -> ChatSession:
+    """
+    This takes a chat session id for a session in Slack and:
+    - Creates a new chat session in the DB
+    - Tries to copy the persona from the original chat session
+        (if it is available to the user clicking the button)
+    - Sets the user to the given user (if provided)
+    """
+    chat_session = get_chat_session_by_id(
+        chat_session_id=chat_session_id,
+        user_id=None,  # Ignore user permissions for this
+        db_session=db_session,
+    )
+    if not chat_session:
+        raise HTTPException(status_code=400, detail="Invalid Chat Session ID provided")
+
+    # This enforces permissions and sets a default
+    new_persona_id = get_best_persona_id_for_user(
+        db_session=db_session,
+        user=user,
+        persona_id=chat_session.persona_id,
+    )
+
+    return create_chat_session(
+        db_session=db_session,
+        user_id=user.id if user else None,
+        persona_id=new_persona_id,
+        # Set this to empty string so the frontend will force a rename
+        description="",
+        llm_override=chat_session.llm_override,
+        prompt_override=chat_session.prompt_override,
+        # Chat is in UI now so this is false
+        danswerbot_flow=False,
+        # Maybe we want this in the future to track if it was created from Slack
+        slack_thread_id=None,
+    )
+
+
 def update_chat_session(
     db_session: Session,
     user_id: UUID | None,
@@ -336,6 +372,28 @@ def get_chat_message(
     return chat_message
 
 
+def get_chat_session_by_message_id(
+    db_session: Session,
+    message_id: int,
+) -> ChatSession:
+    """
+    Should only be used for Slack
+    Get the chat session associated with a specific message ID
+    Note: this ignores permission checks.
+    """
+    stmt = select(ChatMessage).where(ChatMessage.id == message_id)
+
+    result = db_session.execute(stmt)
+    chat_message = result.scalar_one_or_none()
+
+    if chat_message is None:
+        raise ValueError(
+            f"Unable to find chat session associated with message ID: {message_id}"
+        )
+
+    return chat_message.chat_session
+
+
 def get_chat_messages_by_sessions(
     chat_session_ids: list[UUID],
     user_id: UUID | None,
@@ -355,6 +413,44 @@ def get_chat_messages_by_sessions(
     return db_session.execute(stmt).scalars().all()
 
 
+def add_chats_to_session_from_slack_thread(
+    db_session: Session,
+    slack_chat_session_id: UUID,
+    new_chat_session_id: UUID,
+) -> None:
+    new_root_message = get_or_create_root_message(
+        chat_session_id=new_chat_session_id,
+        db_session=db_session,
+    )
+
+    for chat_message in get_chat_messages_by_sessions(
+        chat_session_ids=[slack_chat_session_id],
+        user_id=None,  # Ignore user permissions for this
+        db_session=db_session,
+        skip_permission_check=True,
+    ):
+        if chat_message.message_type == MessageType.SYSTEM:
+            continue
+        # Duplicate the message
+        new_root_message = create_new_chat_message(
+            db_session=db_session,
+            chat_session_id=new_chat_session_id,
+            parent_message=new_root_message,
+            message=chat_message.message,
+            files=chat_message.files,
+            rephrased_query=chat_message.rephrased_query,
+            error=chat_message.error,
+            citations=chat_message.citations,
+            reference_docs=chat_message.search_docs,
+            tool_call=chat_message.tool_call,
+            prompt_id=chat_message.prompt_id,
+            token_count=chat_message.token_count,
+            message_type=chat_message.message_type,
+            alternate_assistant_id=chat_message.alternate_assistant_id,
+            overridden_model=chat_message.overridden_model,
+        )
+
+
 def get_search_docs_for_chat_message(
     chat_message_id: int, db_session: Session
 ) -> list[SearchDoc]:

backend/danswer/db/connector.py

@@ -12,6 +12,7 @@ from sqlalchemy.orm import Session
 from danswer.configs.app_configs import DEFAULT_PRUNING_FREQ
 from danswer.configs.constants import DocumentSource
 from danswer.connectors.models import InputType
+from danswer.db.enums import IndexingMode
 from danswer.db.models import Connector
 from danswer.db.models import ConnectorCredentialPair
 from danswer.db.models import IndexAttempt
@@ -311,3 +312,25 @@ def mark_cc_pair_as_external_group_synced(db_session: Session, cc_pair_id: int)
     # If this changes, we need to update this function.
     cc_pair.last_time_external_group_sync = datetime.now(timezone.utc)
     db_session.commit()
+
+
+def mark_ccpair_with_indexing_trigger(
+    cc_pair_id: int, indexing_mode: IndexingMode | None, db_session: Session
+) -> None:
+    """indexing_mode sets a field which will be picked up by a background task
+    to trigger indexing. Set to None to disable the trigger."""
+    try:
+        cc_pair = db_session.execute(
+            select(ConnectorCredentialPair)
+            .where(ConnectorCredentialPair.id == cc_pair_id)
+            .with_for_update()
+        ).scalar_one()
+
+        if cc_pair is None:
+            raise ValueError(f"No cc_pair with ID: {cc_pair_id}")
+
+        cc_pair.indexing_trigger = indexing_mode
+        db_session.commit()
+    except Exception:
+        db_session.rollback()
+        raise

backend/danswer/db/connector_credential_pair.py

@@ -324,8 +324,11 @@ def associate_default_cc_pair(db_session: Session) -> None:
 def _relate_groups_to_cc_pair__no_commit(
     db_session: Session,
     cc_pair_id: int,
-    user_group_ids: list[int],
+    user_group_ids: list[int] | None = None,
 ) -> None:
+    if not user_group_ids:
+        return
+
     for group_id in user_group_ids:
         db_session.add(
             UserGroup__ConnectorCredentialPair(
@@ -402,12 +405,11 @@ def add_credential_to_connector(
     db_session.flush()  # make sure the association has an id
     db_session.refresh(association)
 
-    if groups and access_type != AccessType.SYNC:
-        _relate_groups_to_cc_pair__no_commit(
-            db_session=db_session,
-            cc_pair_id=association.id,
-            user_group_ids=groups,
-        )
+    _relate_groups_to_cc_pair__no_commit(
+        db_session=db_session,
+        cc_pair_id=association.id,
+        user_group_ids=groups,
+    )
 
     db_session.commit()
 

backend/danswer/db/engine.py

@@ -37,6 +37,7 @@ from danswer.configs.app_configs import POSTGRES_PORT
 from danswer.configs.app_configs import POSTGRES_USER
 from danswer.configs.app_configs import USER_AUTH_SECRET
 from danswer.configs.constants import POSTGRES_UNKNOWN_APP_NAME
+from danswer.server.utils import BasicAuthenticationError
 from danswer.utils.logger import setup_logger
 from shared_configs.configs import MULTI_TENANT
 from shared_configs.configs import POSTGRES_DEFAULT_SCHEMA
@@ -426,7 +427,9 @@ def get_session() -> Generator[Session, None, None]:
     """Generate a database session with the appropriate tenant schema set."""
     tenant_id = CURRENT_TENANT_ID_CONTEXTVAR.get()
     if tenant_id == POSTGRES_DEFAULT_SCHEMA and MULTI_TENANT:
-        raise HTTPException(status_code=401, detail="User must authenticate")
+        raise BasicAuthenticationError(
+            detail="User must authenticate",
+        )
 
     engine = get_sqlalchemy_engine()
 

backend/danswer/db/enums.py

@@ -19,6 +19,11 @@ class IndexingStatus(str, PyEnum):
         return self in terminal_states
 
 
+class IndexingMode(str, PyEnum):
+    UPDATE = "update"
+    REINDEX = "reindex"
+
+
 # these may differ in the future, which is why we're okay with this duplication
 class DeletionStatus(str, PyEnum):
     NOT_STARTED = "not_started"

backend/danswer/db/models.py

@@ -1,6 +1,5 @@
 import datetime
 import json
-from enum import Enum as PyEnum
 from typing import Any
 from typing import Literal
 from typing import NotRequired
@@ -42,7 +41,7 @@ from danswer.configs.constants import DEFAULT_BOOST
 from danswer.configs.constants import DocumentSource
 from danswer.configs.constants import FileOrigin
 from danswer.configs.constants import MessageType
-from danswer.db.enums import AccessType
+from danswer.db.enums import AccessType, IndexingMode
 from danswer.configs.constants import NotificationType
 from danswer.configs.constants import SearchFeedbackType
 from danswer.configs.constants import TokenRateLimitScope
@@ -126,6 +125,7 @@ class User(SQLAlchemyBaseUserTableUUID, Base):
 
     # if specified, controls the assistants that are shown to the user + their order
     # if not specified, all assistants are shown
+    auto_scroll: Mapped[bool] = mapped_column(Boolean, default=True)
     chosen_assistants: Mapped[list[int] | None] = mapped_column(
         postgresql.JSONB(), nullable=True, default=None
     )
@@ -438,6 +438,10 @@ class ConnectorCredentialPair(Base):
 
     total_docs_indexed: Mapped[int] = mapped_column(Integer, default=0)
 
+    indexing_trigger: Mapped[IndexingMode | None] = mapped_column(
+        Enum(IndexingMode, native_enum=False), nullable=True
+    )
+
     connector: Mapped["Connector"] = relationship(
         "Connector", back_populates="credentials"
     )
@@ -959,9 +963,8 @@ class ChatSession(Base):
     persona_id: Mapped[int | None] = mapped_column(
         ForeignKey("persona.id"), nullable=True
     )
-    description: Mapped[str] = mapped_column(Text)
-    # One-shot direct answering, currently the two types of chats are not mixed
-    one_shot: Mapped[bool] = mapped_column(Boolean, default=False)
+    description: Mapped[str | None] = mapped_column(Text, nullable=True)
+    # This chat created by DanswerBot
     danswerbot_flow: Mapped[bool] = mapped_column(Boolean, default=False)
     # Only ever set to True if system is set to not hard-delete chats
     deleted: Mapped[bool] = mapped_column(Boolean, default=False)
@@ -1480,11 +1483,7 @@ class ChannelConfig(TypedDict):
     # If None then no follow up
     # If empty list, follow up with no tags
     follow_up_tags: NotRequired[list[str]]
-
-
-class SlackBotResponseType(str, PyEnum):
-    QUOTES = "quotes"
-    CITATIONS = "citations"
+    show_continue_in_web_ui: NotRequired[bool]  # defaults to False
 
 
 class SlackChannelConfig(Base):
@@ -1499,9 +1498,6 @@ class SlackChannelConfig(Base):
     channel_config: Mapped[ChannelConfig] = mapped_column(
         postgresql.JSONB(), nullable=False
     )
-    response_type: Mapped[SlackBotResponseType] = mapped_column(
-        Enum(SlackBotResponseType, native_enum=False), nullable=False
-    )
 
     enable_auto_filters: Mapped[bool] = mapped_column(
         Boolean, nullable=False, default=False

backend/danswer/db/persona.py

@@ -113,6 +113,31 @@ def fetch_persona_by_id(
     return persona
 
 
+def get_best_persona_id_for_user(
+    db_session: Session, user: User | None, persona_id: int | None = None
+) -> int | None:
+    if persona_id is not None:
+        stmt = select(Persona).where(Persona.id == persona_id).distinct()
+        stmt = _add_user_filters(
+            stmt=stmt,
+            user=user,
+            # We don't want to filter by editable here, we just want to see if the
+            # persona is usable by the user
+            get_editable=False,
+        )
+        persona = db_session.scalars(stmt).one_or_none()
+        if persona:
+            return persona.id
+
+    # If the persona is not found, or the slack bot is using doc sets instead of personas,
+    # we need to find the best persona for the user
+    # This is the persona with the highest display priority that the user has access to
+    stmt = select(Persona).order_by(Persona.display_priority.desc()).distinct()
+    stmt = _add_user_filters(stmt=stmt, user=user, get_editable=True)
+    persona = db_session.scalars(stmt).one_or_none()
+    return persona.id if persona else None
+
+
 def _get_persona_by_name(
     persona_name: str, user: User | None, db_session: Session
 ) -> Persona | None:
@@ -160,7 +185,7 @@ def create_update_persona(
             "persona_id": persona_id,
             "user": user,
             "db_session": db_session,
-            **create_persona_request.dict(exclude={"users", "groups"}),
+            **create_persona_request.model_dump(exclude={"users", "groups"}),
         }
 
         persona = upsert_persona(**persona_data)
@@ -390,9 +415,6 @@ def upsert_prompt(
     return prompt
 
 
-# NOTE: This operation cannot update persona configuration options that
-# are core to the persona, such as its display priority and
-# whether or not the assistant is a built-in / default assistant
 def upsert_persona(
     user: User | None,
     name: str,
@@ -424,6 +446,12 @@ def upsert_persona(
     chunks_above: int = CONTEXT_CHUNKS_ABOVE,
     chunks_below: int = CONTEXT_CHUNKS_BELOW,
 ) -> Persona:
+    """
+    NOTE: This operation cannot update persona configuration options that
+    are core to the persona, such as its display priority and
+    whether or not the assistant is a built-in / default assistant
+    """
+
     if persona_id is not None:
         persona = db_session.query(Persona).filter_by(id=persona_id).first()
     else:
@@ -461,6 +489,8 @@ def upsert_persona(
         validate_persona_tools(tools)
 
     if persona:
+        # Built-in personas can only be updated through YAML configuration.
+        # This ensures that core system personas are not modified unintentionally.
         if persona.builtin_persona and not builtin_persona:
             raise ValueError("Cannot update builtin persona with non-builtin.")
 
@@ -469,6 +499,9 @@ def upsert_persona(
             db_session=db_session, persona_id=persona.id, user=user, get_editable=True
         )
 
+        # The following update excludes `default`, `built-in`, and display priority.
+        # Display priority is handled separately in the `display-priority` endpoint.
+        # `default` and `built-in` properties can only be set when creating a persona.
         persona.name = name
         persona.description = description
         persona.num_chunks = num_chunks
@@ -733,6 +766,8 @@ def get_prompt_by_name(
     if user and user.role != UserRole.ADMIN:
         stmt = stmt.where(Prompt.user_id == user.id)
 
+    # Order by ID to ensure consistent result when multiple prompts exist
+    stmt = stmt.order_by(Prompt.id).limit(1)
     result = db_session.execute(stmt).scalar_one_or_none()
     return result
 

backend/danswer/db/search_settings.py

@@ -143,6 +143,25 @@ def get_secondary_search_settings(db_session: Session) -> SearchSettings | None:
     return latest_settings
 
 
+def get_active_search_settings(db_session: Session) -> list[SearchSettings]:
+    """Returns active search settings. The first entry will always be the current search
+    settings. If there are new search settings that are being migrated to, those will be
+    the second entry."""
+    search_settings_list: list[SearchSettings] = []
+
+    # Get the primary search settings
+    primary_search_settings = get_current_search_settings(db_session)
+    search_settings_list.append(primary_search_settings)
+
+    # Check for secondary search settings
+    secondary_search_settings = get_secondary_search_settings(db_session)
+    if secondary_search_settings is not None:
+        # If secondary settings exist, add them to the list
+        search_settings_list.append(secondary_search_settings)
+
+    return search_settings_list
+
+
 def get_all_search_settings(db_session: Session) -> list[SearchSettings]:
     query = select(SearchSettings).order_by(SearchSettings.id.desc())
     result = db_session.execute(query)

backend/danswer/db/slack_channel_config.py

@@ -10,7 +10,6 @@ from danswer.db.constants import SLACK_BOT_PERSONA_PREFIX
 from danswer.db.models import ChannelConfig
 from danswer.db.models import Persona
 from danswer.db.models import Persona__DocumentSet
-from danswer.db.models import SlackBotResponseType
 from danswer.db.models import SlackChannelConfig
 from danswer.db.models import User
 from danswer.db.persona import get_default_prompt
@@ -83,7 +82,6 @@ def insert_slack_channel_config(
     slack_bot_id: int,
     persona_id: int | None,
     channel_config: ChannelConfig,
-    response_type: SlackBotResponseType,
     standard_answer_category_ids: list[int],
     enable_auto_filters: bool,
 ) -> SlackChannelConfig:
@@ -115,7 +113,6 @@ def insert_slack_channel_config(
         slack_bot_id=slack_bot_id,
         persona_id=persona_id,
         channel_config=channel_config,
-        response_type=response_type,
         standard_answer_categories=existing_standard_answer_categories,
         enable_auto_filters=enable_auto_filters,
     )
@@ -130,7 +127,6 @@ def update_slack_channel_config(
     slack_channel_config_id: int,
     persona_id: int | None,
     channel_config: ChannelConfig,
-    response_type: SlackBotResponseType,
     standard_answer_category_ids: list[int],
     enable_auto_filters: bool,
 ) -> SlackChannelConfig:
@@ -170,7 +166,6 @@ def update_slack_channel_config(
     # will encounter `violates foreign key constraint` errors
     slack_channel_config.persona_id = persona_id
     slack_channel_config.channel_config = channel_config
-    slack_channel_config.response_type = response_type
     slack_channel_config.standard_answer_categories = list(
         existing_standard_answer_categories
     )

backend/danswer/file_processing/extract_file_text.py

@@ -295,7 +295,7 @@ def pptx_to_text(file: IO[Any]) -> str:
 
 
 def xlsx_to_text(file: IO[Any]) -> str:
-    workbook = openpyxl.load_workbook(file)
+    workbook = openpyxl.load_workbook(file, read_only=True)
     text_content = []
     for sheet in workbook.worksheets:
         sheet_string = "\n".join(

backend/danswer/file_store/file_store.py

@@ -59,6 +59,12 @@ class FileStore(ABC):
             Contents of the file and metadata dict
         """
 
+    @abstractmethod
+    def read_file_record(self, file_name: str) -> PGFileStore:
+        """
+        Read the file record by the name
+        """
+
     @abstractmethod
     def delete_file(self, file_name: str) -> None:
         """

backend/danswer/llm/answering/answer.py

@@ -18,18 +18,12 @@ from danswer.llm.answering.models import PromptConfig
 from danswer.llm.answering.prompts.build import AnswerPromptBuilder
 from danswer.llm.answering.prompts.build import default_build_system_message
 from danswer.llm.answering.prompts.build import default_build_user_message
-from danswer.llm.answering.stream_processing.answer_response_handler import (
-    AnswerResponseHandler,
-)
 from danswer.llm.answering.stream_processing.answer_response_handler import (
     CitationResponseHandler,
 )
 from danswer.llm.answering.stream_processing.answer_response_handler import (
     DummyAnswerResponseHandler,
 )
-from danswer.llm.answering.stream_processing.answer_response_handler import (
-    QuotesResponseHandler,
-)
 from danswer.llm.answering.stream_processing.utils import map_document_id_order
 from danswer.llm.answering.tool.tool_response_handler import ToolResponseHandler
 from danswer.llm.interfaces import LLM
@@ -214,18 +208,23 @@ class Answer:
 
         search_result = SearchTool.get_search_result(current_llm_call) or []
 
-        answer_handler: AnswerResponseHandler
-        if self.answer_style_config.citation_config:
-            answer_handler = CitationResponseHandler(
-                context_docs=search_result,
-                doc_id_to_rank_map=map_document_id_order(search_result),
-            )
-        elif self.answer_style_config.quotes_config:
-            answer_handler = QuotesResponseHandler(
-                context_docs=search_result,
-            )
-        else:
-            raise ValueError("No answer style config provided")
+        # Quotes are no longer supported
+        # answer_handler: AnswerResponseHandler
+        # if self.answer_style_config.citation_config:
+        #     answer_handler = CitationResponseHandler(
+        #         context_docs=search_result,
+        #         doc_id_to_rank_map=map_document_id_order(search_result),
+        #     )
+        # elif self.answer_style_config.quotes_config:
+        #     answer_handler = QuotesResponseHandler(
+        #         context_docs=search_result,
+        #     )
+        # else:
+        #     raise ValueError("No answer style config provided")
+        answer_handler = CitationResponseHandler(
+            context_docs=search_result,
+            doc_id_to_rank_map=map_document_id_order(search_result),
+        )
 
         response_handler_manager = LLMResponseHandlerManager(
             tool_call_handler, answer_handler, self.is_cancelled

backend/danswer/llm/answering/llm_response_handler.py

@@ -8,7 +8,6 @@ from pydantic.v1 import BaseModel as BaseModel__v1
 
 from danswer.chat.models import CitationInfo
 from danswer.chat.models import DanswerAnswerPiece
-from danswer.chat.models import DanswerQuotes
 from danswer.chat.models import StreamStopInfo
 from danswer.chat.models import StreamStopReason
 from danswer.file_store.models import InMemoryChatFile
@@ -30,7 +29,6 @@ if TYPE_CHECKING:
 ResponsePart = (
     DanswerAnswerPiece
     | CitationInfo
-    | DanswerQuotes
     | ToolCallKickoff
     | ToolResponse
     | ToolCallFinalResult

backend/danswer/llm/answering/stream_processing/answer_response_handler.py

@@ -9,9 +9,6 @@ from danswer.llm.answering.llm_response_handler import ResponsePart
 from danswer.llm.answering.stream_processing.citation_processing import (
     CitationProcessor,
 )
-from danswer.llm.answering.stream_processing.quotes_processing import (
-    QuotesProcessor,
-)
 from danswer.llm.answering.stream_processing.utils import DocumentIdOrderMapping
 from danswer.utils.logger import setup_logger
 
@@ -70,28 +67,29 @@ class CitationResponseHandler(AnswerResponseHandler):
         yield from self.citation_processor.process_token(content)
 
 
-class QuotesResponseHandler(AnswerResponseHandler):
-    def __init__(
-        self,
-        context_docs: list[LlmDoc],
-        is_json_prompt: bool = True,
-    ):
-        self.quotes_processor = QuotesProcessor(
-            context_docs=context_docs,
-            is_json_prompt=is_json_prompt,
-        )
+# No longer in use, remove later
+# class QuotesResponseHandler(AnswerResponseHandler):
+#     def __init__(
+#         self,
+#         context_docs: list[LlmDoc],
+#         is_json_prompt: bool = True,
+#     ):
+#         self.quotes_processor = QuotesProcessor(
+#             context_docs=context_docs,
+#             is_json_prompt=is_json_prompt,
+#         )
 
-    def handle_response_part(
-        self,
-        response_item: BaseMessage | None,
-        previous_response_items: list[BaseMessage],
-    ) -> Generator[ResponsePart, None, None]:
-        if response_item is None:
-            yield from self.quotes_processor.process_token(None)
-            return
+#     def handle_response_part(
+#         self,
+#         response_item: BaseMessage | None,
+#         previous_response_items: list[BaseMessage],
+#     ) -> Generator[ResponsePart, None, None]:
+#         if response_item is None:
+#             yield from self.quotes_processor.process_token(None)
+#             return
 
-        content = (
-            response_item.content if isinstance(response_item.content, str) else ""
-        )
+#         content = (
+#             response_item.content if isinstance(response_item.content, str) else ""
+#         )
 
-        yield from self.quotes_processor.process_token(content)
+#         yield from self.quotes_processor.process_token(content)

backend/danswer/llm/answering/stream_processing/citation_processing.py

@@ -67,9 +67,9 @@ class CitationProcessor:
                 if piece_that_comes_after == "\n" and in_code_block(self.llm_out):
                     self.curr_segment = self.curr_segment.replace("```", "```plaintext")
 
-        citation_pattern = r"\[(\d+)\]"
+        citation_pattern = r"\[(\d+)\]|\[\[(\d+)\]\]"  # [1], [[1]], etc.
         citations_found = list(re.finditer(citation_pattern, self.curr_segment))
-        possible_citation_pattern = r"(\[\d*$)"  # [1, [, etc
+        possible_citation_pattern = r"(\[+\d*$)"  # [1, [, [[, [[2, etc.
         possible_citation_found = re.search(
             possible_citation_pattern, self.curr_segment
         )
@@ -77,13 +77,15 @@ class CitationProcessor:
         if len(citations_found) == 0 and len(self.llm_out) - self.past_cite_count > 5:
             self.current_citations = []
 
-        result = ""  # Initialize result here
+        result = ""
         if citations_found and not in_code_block(self.llm_out):
             last_citation_end = 0
             length_to_add = 0
             while len(citations_found) > 0:
                 citation = citations_found.pop(0)
-                numerical_value = int(citation.group(1))
+                numerical_value = int(
+                    next(group for group in citation.groups() if group is not None)
+                )
 
                 if 1 <= numerical_value <= self.max_citation_num:
                     context_llm_doc = self.context_docs[numerical_value - 1]
@@ -131,14 +133,6 @@ class CitationProcessor:
 
                     link = context_llm_doc.link
 
-                    # Replace the citation in the current segment
-                    start, end = citation.span()
-                    self.curr_segment = (
-                        self.curr_segment[: start + length_to_add]
-                        + f"[{target_citation_num}]"
-                        + self.curr_segment[end + length_to_add :]
-                    )
-
                     self.past_cite_count = len(self.llm_out)
                     self.current_citations.append(target_citation_num)
 
@@ -149,6 +143,7 @@ class CitationProcessor:
                             document_id=context_llm_doc.document_id,
                         )
 
+                    start, end = citation.span()
                     if link:
                         prev_length = len(self.curr_segment)
                         self.curr_segment = (

backend/danswer/llm/answering/stream_processing/quotes_processing.py

@@ -1,3 +1,4 @@
+# THIS IS NO LONGER IN USE
 import math
 import re
 from collections.abc import Generator
@@ -5,11 +6,10 @@ from json import JSONDecodeError
 from typing import Optional
 
 import regex
+from pydantic import BaseModel
 
 from danswer.chat.models import DanswerAnswer
 from danswer.chat.models import DanswerAnswerPiece
-from danswer.chat.models import DanswerQuote
-from danswer.chat.models import DanswerQuotes
 from danswer.chat.models import LlmDoc
 from danswer.configs.chat_configs import QUOTE_ALLOWED_ERROR_PERCENT
 from danswer.context.search.models import InferenceChunk
@@ -26,6 +26,20 @@ logger = setup_logger()
 answer_pattern = re.compile(r'{\s*"answer"\s*:\s*"', re.IGNORECASE)
 
 
+class DanswerQuote(BaseModel):
+    # This is during inference so everything is a string by this point
+    quote: str
+    document_id: str
+    link: str | None
+    source_type: str
+    semantic_identifier: str
+    blurb: str
+
+
+class DanswerQuotes(BaseModel):
+    quotes: list[DanswerQuote]
+
+
 def _extract_answer_quotes_freeform(
     answer_raw: str,
 ) -> tuple[Optional[str], Optional[list[str]]]:

backend/danswer/llm/chat_llm.py

@@ -26,7 +26,9 @@ from langchain_core.messages.tool import ToolMessage
 from langchain_core.prompt_values import PromptValue
 
 from danswer.configs.app_configs import LOG_DANSWER_MODEL_INTERACTIONS
-from danswer.configs.model_configs import DISABLE_LITELLM_STREAMING
+from danswer.configs.model_configs import (
+    DISABLE_LITELLM_STREAMING,
+)
 from danswer.configs.model_configs import GEN_AI_TEMPERATURE
 from danswer.configs.model_configs import LITELLM_EXTRA_BODY
 from danswer.llm.interfaces import LLM
@@ -161,7 +163,9 @@ def _convert_delta_to_message_chunk(
 
     if role == "user":
         return HumanMessageChunk(content=content)
-    elif role == "assistant":
+    # NOTE: if tool calls are present, then it's an assistant.
+    # In Ollama, the role will be None for tool-calls
+    elif role == "assistant" or tool_calls:
         if tool_calls:
             tool_call = tool_calls[0]
             tool_name = tool_call.function.name or (curr_msg and curr_msg.name) or ""
@@ -236,6 +240,7 @@ class DefaultMultiLLM(LLM):
         custom_config: dict[str, str] | None = None,
         extra_headers: dict[str, str] | None = None,
         extra_body: dict | None = LITELLM_EXTRA_BODY,
+        model_kwargs: dict[str, Any] | None = None,
         long_term_logger: LongTermLogger | None = None,
     ):
         self._timeout = timeout
@@ -268,7 +273,7 @@ class DefaultMultiLLM(LLM):
             for k, v in custom_config.items():
                 os.environ[k] = v
 
-        model_kwargs: dict[str, Any] = {}
+        model_kwargs = model_kwargs or {}
         if extra_headers:
             model_kwargs.update({"extra_headers": extra_headers})
         if extra_body:

backend/danswer/llm/factory.py

@@ -1,5 +1,9 @@
+from typing import Any
+
+from danswer.chat.models import PersonaOverrideConfig
 from danswer.configs.app_configs import DISABLE_GENERATIVE_AI
 from danswer.configs.chat_configs import QA_TIMEOUT
+from danswer.configs.model_configs import GEN_AI_MODEL_FALLBACK_MAX_TOKENS
 from danswer.configs.model_configs import GEN_AI_TEMPERATURE
 from danswer.db.engine import get_session_context_manager
 from danswer.db.llm import fetch_default_provider
@@ -10,8 +14,20 @@ from danswer.llm.exceptions import GenAIDisabledException
 from danswer.llm.interfaces import LLM
 from danswer.llm.override_models import LLMOverride
 from danswer.utils.headers import build_llm_extra_headers
+from danswer.utils.logger import setup_logger
 from danswer.utils.long_term_log import LongTermLogger
 
+logger = setup_logger()
+
+
+def _build_extra_model_kwargs(provider: str) -> dict[str, Any]:
+    """Ollama requires us to specify the max context window.
+
+    For now, just using the GEN_AI_MODEL_FALLBACK_MAX_TOKENS value.
+    TODO: allow model-specific values to be configured via the UI.
+    """
+    return {"num_ctx": GEN_AI_MODEL_FALLBACK_MAX_TOKENS} if provider == "ollama" else {}
+
 
 def get_main_llm_from_tuple(
     llms: tuple[LLM, LLM],
@@ -20,11 +36,15 @@ def get_main_llm_from_tuple(
 
 
 def get_llms_for_persona(
-    persona: Persona,
+    persona: Persona | PersonaOverrideConfig | None,
     llm_override: LLMOverride | None = None,
     additional_headers: dict[str, str] | None = None,
     long_term_logger: LongTermLogger | None = None,
 ) -> tuple[LLM, LLM]:
+    if persona is None:
+        logger.warning("No persona provided, using default LLMs")
+        return get_default_llms()
+
     model_provider_override = llm_override.model_provider if llm_override else None
     model_version_override = llm_override.model_version if llm_override else None
     temperature_override = llm_override.temperature if llm_override else None
@@ -59,6 +79,7 @@ def get_llms_for_persona(
             api_base=llm_provider.api_base,
             api_version=llm_provider.api_version,
             custom_config=llm_provider.custom_config,
+            temperature=temperature_override,
             additional_headers=additional_headers,
             long_term_logger=long_term_logger,
         )
@@ -116,11 +137,13 @@ def get_llm(
     api_base: str | None = None,
     api_version: str | None = None,
     custom_config: dict[str, str] | None = None,
-    temperature: float = GEN_AI_TEMPERATURE,
+    temperature: float | None = None,
     timeout: int = QA_TIMEOUT,
     additional_headers: dict[str, str] | None = None,
     long_term_logger: LongTermLogger | None = None,
 ) -> LLM:
+    if temperature is None:
+        temperature = GEN_AI_TEMPERATURE
     return DefaultMultiLLM(
         model_provider=provider,
         model_name=model,
@@ -132,5 +155,6 @@ def get_llm(
         temperature=temperature,
         custom_config=custom_config,
         extra_headers=build_llm_extra_headers(additional_headers),
+        model_kwargs=_build_extra_model_kwargs(provider),
         long_term_logger=long_term_logger,
     )

backend/danswer/llm/utils.py

@@ -1,3 +1,4 @@
+import copy
 import io
 import json
 from collections.abc import Callable
@@ -385,6 +386,62 @@ def test_llm(llm: LLM) -> str | None:
     return error_msg
 
 
+def get_model_map() -> dict:
+    starting_map = copy.deepcopy(cast(dict, litellm.model_cost))
+
+    # NOTE: we could add additional models here in the future,
+    # but for now there is no point. Ollama allows the user to
+    # to specify their desired max context window, and it's
+    # unlikely to be standard across users even for the same model
+    # (it heavily depends on their hardware). For now, we'll just
+    # rely on GEN_AI_MODEL_FALLBACK_MAX_TOKENS to cover this.
+    # for model_name in [
+    #     "llama3.2",
+    #     "llama3.2:1b",
+    #     "llama3.2:3b",
+    #     "llama3.2:11b",
+    #     "llama3.2:90b",
+    # ]:
+    #     starting_map[f"ollama/{model_name}"] = {
+    #         "max_tokens": 128000,
+    #         "max_input_tokens": 128000,
+    #         "max_output_tokens": 128000,
+    #     }
+
+    return starting_map
+
+
+def _strip_extra_provider_from_model_name(model_name: str) -> str:
+    return model_name.split("/")[1] if "/" in model_name else model_name
+
+
+def _strip_colon_from_model_name(model_name: str) -> str:
+    return ":".join(model_name.split(":")[:-1]) if ":" in model_name else model_name
+
+
+def _find_model_obj(
+    model_map: dict, provider: str, model_names: list[str | None]
+) -> dict | None:
+    # Filter out None values and deduplicate model names
+    filtered_model_names = [name for name in model_names if name]
+
+    # First try all model names with provider prefix
+    for model_name in filtered_model_names:
+        model_obj = model_map.get(f"{provider}/{model_name}")
+        if model_obj:
+            logger.debug(f"Using model object for {provider}/{model_name}")
+            return model_obj
+
+    # Then try all model names without provider prefix
+    for model_name in filtered_model_names:
+        model_obj = model_map.get(model_name)
+        if model_obj:
+            logger.debug(f"Using model object for {model_name}")
+            return model_obj
+
+    return None
+
+
 def get_llm_max_tokens(
     model_map: dict,
     model_name: str,
@@ -397,22 +454,22 @@ def get_llm_max_tokens(
         return GEN_AI_MAX_TOKENS
 
     try:
-        model_obj = model_map.get(f"{model_provider}/{model_name}")
-        if model_obj:
-            logger.debug(f"Using model object for {model_provider}/{model_name}")
-
-        if not model_obj:
-            model_obj = model_map.get(model_name)
-            if model_obj:
-                logger.debug(f"Using model object for {model_name}")
-
-        if not model_obj:
-            model_name_split = model_name.split("/")
-            if len(model_name_split) > 1:
-                model_obj = model_map.get(model_name_split[1])
-            if model_obj:
-                logger.debug(f"Using model object for {model_name_split[1]}")
-
+        extra_provider_stripped_model_name = _strip_extra_provider_from_model_name(
+            model_name
+        )
+        model_obj = _find_model_obj(
+            model_map,
+            model_provider,
+            [
+                model_name,
+                # Remove leading extra provider. Usually for cases where user has a
+                # customer model proxy which appends another prefix
+                extra_provider_stripped_model_name,
+                # remove :XXXX from the end, if present. Needed for ollama.
+                _strip_colon_from_model_name(model_name),
+                _strip_colon_from_model_name(extra_provider_stripped_model_name),
+            ],
+        )
         if not model_obj:
             raise RuntimeError(
                 f"No litellm entry found for {model_provider}/{model_name}"
@@ -488,7 +545,7 @@ def get_max_input_tokens(
     # `model_cost` dict is a named public interface:
     # https://litellm.vercel.app/docs/completion/token_usage#7-model_cost
     # model_map is  litellm.model_cost
-    litellm_model_map = litellm.model_cost
+    litellm_model_map = get_model_map()
 
     input_toks = (
         get_llm_max_tokens(

backend/danswer/main.py

@@ -25,7 +25,7 @@ from danswer.auth.schemas import UserCreate
 from danswer.auth.schemas import UserRead
 from danswer.auth.schemas import UserUpdate
 from danswer.auth.users import auth_backend
-from danswer.auth.users import BasicAuthenticationError
+from danswer.auth.users import create_danswer_oauth_router
 from danswer.auth.users import fastapi_users
 from danswer.configs.app_configs import APP_API_PREFIX
 from danswer.configs.app_configs import APP_HOST
@@ -44,6 +44,7 @@ from danswer.configs.constants import AuthType
 from danswer.configs.constants import POSTGRES_WEB_APP_NAME
 from danswer.db.engine import SqlEngine
 from danswer.db.engine import warm_up_connections
+from danswer.server.api_key.api import router as api_key_router
 from danswer.server.auth_check import check_router_auth
 from danswer.server.danswer_api.ingestion import router as danswer_api_router
 from danswer.server.documents.cc_pair import router as cc_pair_router
@@ -90,6 +91,7 @@ from danswer.server.settings.api import basic_router as settings_router
 from danswer.server.token_rate_limits.api import (
     router as token_rate_limit_settings_router,
 )
+from danswer.server.utils import BasicAuthenticationError
 from danswer.setup import setup_danswer
 from danswer.setup import setup_multitenant_danswer
 from danswer.utils.logger import setup_logger
@@ -204,7 +206,7 @@ def log_http_error(_: Request, exc: Exception) -> JSONResponse:
 
     if isinstance(exc, BasicAuthenticationError):
         # For BasicAuthenticationError, just log a brief message without stack trace (almost always spam)
-        logger.error(f"Authentication failed: {str(exc)}")
+        logger.warning(f"Authentication failed: {str(exc)}")
 
     elif status_code >= 400:
         error_msg = f"{str(exc)}\n"
@@ -280,6 +282,7 @@ def get_application() -> FastAPI:
         application, get_full_openai_assistants_api_router()
     )
     include_router_with_global_prefix_prepended(application, long_term_logs_router)
+    include_router_with_global_prefix_prepended(application, api_key_router)
 
     if AUTH_TYPE == AuthType.DISABLED:
         # Server logs this during auth setup verification step
@@ -323,7 +326,7 @@ def get_application() -> FastAPI:
         oauth_client = GoogleOAuth2(OAUTH_CLIENT_ID, OAUTH_CLIENT_SECRET)
         include_router_with_global_prefix_prepended(
             application,
-            fastapi_users.get_oauth_router(
+            create_danswer_oauth_router(
                 oauth_client,
                 auth_backend,
                 USER_AUTH_SECRET,

backend/danswer/natural_language_processing/exceptions.py

@@ -0,0 +1,4 @@
+class ModelServerRateLimitError(Exception):
+    """
+    Exception raised for rate limiting errors from the model server.
+    """

backend/danswer/natural_language_processing/search_nlp_models.py

@@ -6,6 +6,9 @@ from typing import Any
 
 import requests
 from httpx import HTTPError
+from requests import JSONDecodeError
+from requests import RequestException
+from requests import Response
 from retry import retry
 
 from danswer.configs.app_configs import LARGE_CHUNK_RATIO
@@ -16,6 +19,9 @@ from danswer.configs.model_configs import (
 from danswer.configs.model_configs import DOC_EMBEDDING_CONTEXT_SIZE
 from danswer.db.models import SearchSettings
 from danswer.indexing.indexing_heartbeat import IndexingHeartbeatInterface
+from danswer.natural_language_processing.exceptions import (
+    ModelServerRateLimitError,
+)
 from danswer.natural_language_processing.utils import get_tokenizer
 from danswer.natural_language_processing.utils import tokenizer_trim_content
 from danswer.utils.logger import setup_logger
@@ -99,28 +105,43 @@ class EmbeddingModel:
         self.embed_server_endpoint = f"{model_server_url}/encoder/bi-encoder-embed"
 
     def _make_model_server_request(self, embed_request: EmbedRequest) -> EmbedResponse:
-        def _make_request() -> EmbedResponse:
+        def _make_request() -> Response:
             response = requests.post(
                 self.embed_server_endpoint, json=embed_request.model_dump()
             )
-            try:
-                response.raise_for_status()
-            except requests.HTTPError as e:
-                try:
-                    error_detail = response.json().get("detail", str(e))
-                except Exception:
-                    error_detail = response.text
-                raise HTTPError(f"HTTP error occurred: {error_detail}") from e
-            except requests.RequestException as e:
-                raise HTTPError(f"Request failed: {str(e)}") from e
+            # signify that this is a rate limit error
+            if response.status_code == 429:
+                raise ModelServerRateLimitError(response.text)
 
-            return EmbedResponse(**response.json())
+            response.raise_for_status()
+            return response
 
-        # only perform retries for the non-realtime embedding of passages (e.g. for indexing)
+        final_make_request_func = _make_request
+
+        # if the text type is a passage, add some default
+        # retries + handling for rate limiting
         if embed_request.text_type == EmbedTextType.PASSAGE:
-            return retry(tries=3, delay=5)(_make_request)()
-        else:
-            return _make_request()
+            final_make_request_func = retry(
+                tries=3,
+                delay=5,
+                exceptions=(RequestException, ValueError, JSONDecodeError),
+            )(final_make_request_func)
+            # use 10 second delay as per Azure suggestion
+            final_make_request_func = retry(
+                tries=10, delay=10, exceptions=ModelServerRateLimitError
+            )(final_make_request_func)
+
+        try:
+            response = final_make_request_func()
+            return EmbedResponse(**response.json())
+        except requests.HTTPError as e:
+            try:
+                error_detail = response.json().get("detail", str(e))
+            except Exception:
+                error_detail = response.text
+            raise HTTPError(f"HTTP error occurred: {error_detail}") from e
+        except requests.RequestException as e:
+            raise HTTPError(f"Request failed: {str(e)}") from e
 
     def _batch_encode_texts(
         self,