Better Logging (#184)

CONTRIBUTING.md

@@ -125,7 +125,7 @@ The first time running Danswer, you will need to run the migrations. Navigate to
 alembic upgrade head
 ```
 
-To run the backend api server, navigate to `danswer/backend` and run:
+To run the backend API server, navigate to `danswer/backend` and run:
 ```bash
 DISABLE_AUTH=True TYPESENSE_API_KEY=local_dev_typesense DYNAMIC_CONFIG_DIR_PATH=./dynamic_config_storage uvicorn danswer.main:app --reload --port 8080
 ```
@@ -135,6 +135,7 @@ To run the background job to check for connector updates and index documents, na
 PYTHONPATH=. TYPESENSE_API_KEY=local_dev_typesense DYNAMIC_CONFIG_DIR_PATH=./dynamic_config_storage python danswer/background/update.py
 ```
 
+Note: if you need finer logging, add the additional environment variable `LOG_LEVEL=DEBUG` to the relevant services.
 
 ### Formatting and Linting
 #### Backend

backend/danswer/auth/users.py

@@ -27,7 +27,7 @@ from danswer.db.auth import get_user_db
 from danswer.db.engine import get_async_session
 from danswer.db.models import AccessToken
 from danswer.db.models import User
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from fastapi import Depends
 from fastapi import HTTPException
 from fastapi import Request

backend/danswer/background/update.py

@@ -22,7 +22,7 @@ from danswer.db.models import Connector
 from danswer.db.models import IndexAttempt
 from danswer.db.models import IndexingStatus
 from danswer.utils.indexing_pipeline import build_indexing_pipeline
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from sqlalchemy.orm import Session
 
 logger = setup_logger()

backend/danswer/background/utils.py

@@ -2,7 +2,7 @@ import time
 from collections.abc import Callable
 from typing import Any
 
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 
 
 logger = setup_logger()

backend/danswer/chunking/models.py

@@ -56,4 +56,6 @@ class InferenceChunk(BaseChunk):
             }
         if METADATA in init_kwargs:
             init_kwargs[METADATA] = json.loads(init_kwargs[METADATA])
+        else:
+            init_kwargs[METADATA] = {}
         return cls(**init_kwargs)

backend/danswer/configs/app_configs.py

@@ -138,6 +138,8 @@ DYNAMIC_CONFIG_STORE = os.environ.get(
     "DYNAMIC_CONFIG_STORE", "FileSystemBackedDynamicConfigStore"
 )
 DYNAMIC_CONFIG_DIR_PATH = os.environ.get("DYNAMIC_CONFIG_DIR_PATH", "/home/storage")
+# notset, debug, info, warning, error, or critical
+LOG_LEVEL = os.environ.get("LOG_LEVEL", "info")
 
 
 #####

backend/danswer/connectors/danswer_jira/connector.py

@@ -11,7 +11,7 @@ from danswer.connectors.interfaces import PollConnector
 from danswer.connectors.interfaces import SecondsSinceUnixEpoch
 from danswer.connectors.models import Document
 from danswer.connectors.models import Section
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from jira import JIRA
 from jira.resources import Issue
 

backend/danswer/connectors/file/connector.py

@@ -15,7 +15,7 @@ from danswer.connectors.interfaces import GenerateDocumentsOutput
 from danswer.connectors.interfaces import LoadConnector
 from danswer.connectors.models import Document
 from danswer.connectors.models import Section
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 
 
 logger = setup_logger()

backend/danswer/connectors/github/connector.py

@@ -8,7 +8,7 @@ from danswer.connectors.interfaces import GenerateDocumentsOutput
 from danswer.connectors.interfaces import LoadConnector
 from danswer.connectors.models import Document
 from danswer.connectors.models import Section
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from github import Github
 from github.PaginatedList import PaginatedList
 from github.PullRequest import PullRequest

backend/danswer/connectors/google_drive/connector.py

@@ -14,7 +14,7 @@ from danswer.connectors.interfaces import PollConnector
 from danswer.connectors.interfaces import SecondsSinceUnixEpoch
 from danswer.connectors.models import Document
 from danswer.connectors.models import Section
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from google.auth.transport.requests import Request  # type: ignore
 from google.oauth2.credentials import Credentials  # type: ignore
 from google_auth_oauthlib.flow import InstalledAppFlow  # type: ignore

backend/danswer/connectors/google_drive/connector_auth.py

@@ -9,7 +9,7 @@ from danswer.db.credentials import update_credential_json
 from danswer.db.models import User
 from danswer.dynamic_configs import get_dynamic_config_store
 from danswer.server.models import GoogleAppCredentials
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from google.auth.transport.requests import Request  # type: ignore
 from google.oauth2.credentials import Credentials  # type: ignore
 from google_auth_oauthlib.flow import InstalledAppFlow  # type: ignore

backend/danswer/connectors/slab/connector.py

@@ -15,7 +15,7 @@ from danswer.connectors.interfaces import PollConnector
 from danswer.connectors.interfaces import SecondsSinceUnixEpoch
 from danswer.connectors.models import Document
 from danswer.connectors.models import Section
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from dateutil import parser
 
 # Fairly generous retry because it's not understood why occasionally GraphQL requests fail even with timeout > 1 min

backend/danswer/connectors/slack/connector.py

@@ -18,7 +18,7 @@ from danswer.connectors.slack.utils import get_message_link
 from danswer.connectors.slack.utils import make_slack_api_call_paginated
 from danswer.connectors.slack.utils import make_slack_api_rate_limited
 from danswer.connectors.slack.utils import UserIdReplacer
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from slack_sdk import WebClient
 from slack_sdk.web import SlackResponse
 

backend/danswer/connectors/slack/utils.py

@@ -4,7 +4,7 @@ from collections.abc import Callable
 from typing import Any
 from typing import cast
 
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from slack_sdk import WebClient
 from slack_sdk.errors import SlackApiError
 from slack_sdk.web import SlackResponse

backend/danswer/connectors/web/connector.py

@@ -14,7 +14,7 @@ from danswer.connectors.interfaces import GenerateDocumentsOutput
 from danswer.connectors.interfaces import LoadConnector
 from danswer.connectors.models import Document
 from danswer.connectors.models import Section
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from playwright.sync_api import sync_playwright
 from PyPDF2 import PdfReader
 

backend/danswer/datastores/qdrant/indexing.py

@@ -20,7 +20,7 @@ from danswer.datastores.datastore_utils import DEFAULT_BATCH_SIZE
 from danswer.datastores.datastore_utils import get_uuid_from_chunk
 from danswer.datastores.datastore_utils import update_doc_user_map
 from danswer.utils.clients import get_qdrant_client
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from qdrant_client import QdrantClient
 from qdrant_client.http import models
 from qdrant_client.http.exceptions import ResponseHandlingException

backend/danswer/datastores/qdrant/store.py

@@ -13,7 +13,7 @@ from danswer.datastores.interfaces import VectorIndex
 from danswer.datastores.qdrant.indexing import index_qdrant_chunks
 from danswer.search.search_utils import get_default_embedding_model
 from danswer.utils.clients import get_qdrant_client
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from danswer.utils.timing import log_function_time
 from qdrant_client.http.exceptions import ResponseHandlingException
 from qdrant_client.http.exceptions import UnexpectedResponse

backend/danswer/datastores/typesense/store.py

@@ -26,7 +26,7 @@ from danswer.datastores.datastore_utils import update_doc_user_map
 from danswer.datastores.interfaces import IndexFilter
 from danswer.datastores.interfaces import KeywordIndex
 from danswer.utils.clients import get_typesense_client
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from typesense.exceptions import ObjectNotFound  # type: ignore
 
 

backend/danswer/db/connector.py

@@ -7,7 +7,7 @@ from danswer.db.models import IndexAttempt
 from danswer.server.models import ConnectorBase
 from danswer.server.models import ObjectCreationIdResponse
 from danswer.server.models import StatusResponse
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from fastapi import HTTPException
 from sqlalchemy import and_
 from sqlalchemy import func

backend/danswer/db/connector_credential_pair.py

@@ -4,7 +4,7 @@ from danswer.db.models import ConnectorCredentialPair
 from danswer.db.models import IndexingStatus
 from danswer.db.models import User
 from danswer.server.models import StatusResponse
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from fastapi import HTTPException
 from sqlalchemy import func
 from sqlalchemy import select

backend/danswer/db/credentials.py

@@ -5,7 +5,7 @@ from danswer.db.models import Credential
 from danswer.db.models import User
 from danswer.server.models import CredentialBase
 from danswer.server.models import ObjectCreationIdResponse
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from sqlalchemy import select
 from sqlalchemy.orm import Session
 from sqlalchemy.sql.expression import or_

backend/danswer/db/index_attempt.py

@@ -1,7 +1,7 @@
 from danswer.db.engine import translate_db_time_to_server_time
 from danswer.db.models import IndexAttempt
 from danswer.db.models import IndexingStatus
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from sqlalchemy import desc
 from sqlalchemy import select
 from sqlalchemy.orm import Session

backend/danswer/direct_qa/answer_question.py

@@ -14,7 +14,7 @@ from danswer.search.semantic_search import chunks_to_search_docs
 from danswer.search.semantic_search import retrieve_ranked_documents
 from danswer.server.models import QAResponse
 from danswer.server.models import QuestionRequest
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 
 logger = setup_logger()
 

backend/danswer/direct_qa/llm.py

@@ -35,7 +35,8 @@ from danswer.direct_qa.qa_prompts import json_processor
 from danswer.direct_qa.qa_prompts import QUOTE_PAT
 from danswer.direct_qa.qa_prompts import UNCERTAINTY_PAT
 from danswer.dynamic_configs import get_dynamic_config_store
-from danswer.utils.logging import setup_logger
+from danswer.dynamic_configs.interface import ConfigNotFoundError
+from danswer.utils.logger import setup_logger
 from danswer.utils.text_processing import clean_model_quote
 from danswer.utils.text_processing import shared_precompare_cleanup
 from danswer.utils.timing import log_function_time
@@ -179,11 +180,20 @@ def process_answer(
 ) -> tuple[str | None, dict[str, dict[str, str | int | None]] | None]:
     answer, quote_strings = separate_answer_quotes(answer_raw)
     if answer == UNCERTAINTY_PAT or not answer:
+        if answer == UNCERTAINTY_PAT:
+            logger.debug("Answer matched UNCERTAINTY_PAT")
+        else:
+            logger.debug("No answer extracted from raw output")
         return None, None
 
+    logger.info(f"Answer: {answer}")
     if not quote_strings:
+        logger.debug("No quotes extracted from raw output")
         return answer, None
+    logger.info(f"All quotes (including unmatched): {quote_strings}")
     quotes_dict = match_quotes_to_docs(quote_strings, chunks)
+    logger.info(f"Final quotes dict: {quotes_dict}")
+
     return answer, quotes_dict
 
 
@@ -263,9 +273,12 @@ class OpenAICompletionQA(OpenAIQAModel):
         self.prompt_processor = prompt_processor
         self.model_version = model_version
         self.max_output_tokens = max_output_tokens
-        self.api_key = api_key or get_openai_api_key()
         self.timeout = timeout
         self.include_metadata = include_metadata
+        try:
+            self.api_key = api_key or get_openai_api_key()
+        except ConfigNotFoundError:
+            raise RuntimeError("No OpenAI Key available")
 
     @log_function_time()
     def answer_question(
@@ -373,9 +386,12 @@ class OpenAIChatCompletionQA(OpenAIQAModel):
         self.model_version = model_version
         self.max_output_tokens = max_output_tokens
         self.reflexion_try_count = reflexion_try_count
-        self.api_key = api_key or get_openai_api_key()
         self.timeout = timeout
         self.include_metadata = include_metadata
+        try:
+            self.api_key = api_key or get_openai_api_key()
+        except ConfigNotFoundError:
+            raise RuntimeError("No OpenAI Key available")
 
     @log_function_time()
     def answer_question(
@@ -384,7 +400,7 @@ class OpenAIChatCompletionQA(OpenAIQAModel):
         context_docs: list[InferenceChunk],
     ) -> tuple[str | None, dict[str, dict[str, str | int | None]] | None]:
         messages = self.prompt_processor(query, context_docs, self.include_metadata)
-        logger.debug(messages)
+        logger.debug(json.dumps(messages, indent=4))
         model_output = ""
         for _ in range(self.reflexion_try_count + 1):
             openai_call = _handle_openai_exceptions_wrapper(
@@ -418,7 +434,7 @@ class OpenAIChatCompletionQA(OpenAIQAModel):
         self, query: str, context_docs: list[InferenceChunk]
     ) -> Generator[dict[str, Any] | None, None, None]:
         messages = self.prompt_processor(query, context_docs, self.include_metadata)
-        logger.debug(messages)
+        logger.debug(json.dumps(messages, indent=4))
 
         openai_call = _handle_openai_exceptions_wrapper(
             openai_call=openai.ChatCompletion.create,
@@ -446,10 +462,14 @@ class OpenAIChatCompletionQA(OpenAIQAModel):
             event_text = event_dict["content"]
             model_previous = model_output
             model_output += event_text
+            logger.debug(f"GPT returned token: {event_text}")
 
             if not found_answer_start and '{"answer":"' in model_output.replace(
                 " ", ""
             ).replace("\n", ""):
+                # Note, if the token that completes the pattern has additional text, for example if the token is "?
+                # Then the chars after " will not be streamed, but this is ok as it prevents streaming the ? in the
+                # event that the model outputs the UNCERTAINTY_PAT
                 found_answer_start = True
                 continue
 
@@ -463,14 +483,5 @@ class OpenAIChatCompletionQA(OpenAIQAModel):
         logger.debug(model_output)
 
         answer, quotes_dict = process_answer(model_output, context_docs)
-        if answer:
-            logger.info(answer)
-        else:
-            logger.warning(
-                "Answer extraction from model output failed, most likely no quotes provided"
-            )
 
-        if quotes_dict is None:
-            yield {}
-        else:
-            yield quotes_dict
+        yield {} if quotes_dict is None else quotes_dict

backend/danswer/listeners/slack_listener.py

@@ -9,7 +9,7 @@ from danswer.direct_qa.answer_question import answer_question
 from danswer.server.models import QAResponse
 from danswer.server.models import QuestionRequest
 from danswer.server.models import SearchDoc
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from retry import retry
 from slack_sdk import WebClient
 from slack_sdk.socket_mode import SocketModeClient

backend/danswer/main.py

@@ -8,7 +8,10 @@ from danswer.auth.users import fastapi_users
 from danswer.auth.users import google_oauth_client
 from danswer.configs.app_configs import APP_HOST
 from danswer.configs.app_configs import APP_PORT
+from danswer.configs.app_configs import DISABLE_AUTH
 from danswer.configs.app_configs import ENABLE_OAUTH
+from danswer.configs.app_configs import GOOGLE_OAUTH_CLIENT_ID
+from danswer.configs.app_configs import GOOGLE_OAUTH_CLIENT_SECRET
 from danswer.configs.app_configs import SECRET
 from danswer.configs.app_configs import TYPESENSE_DEFAULT_COLLECTION
 from danswer.configs.app_configs import WEB_DOMAIN
@@ -16,11 +19,14 @@ from danswer.datastores.qdrant.indexing import list_qdrant_collections
 from danswer.datastores.typesense.store import check_typesense_collection_exist
 from danswer.datastores.typesense.store import create_typesense_collection
 from danswer.db.credentials import create_initial_public_credential
+from danswer.direct_qa.key_validation import check_openai_api_key_is_valid
+from danswer.direct_qa.llm import get_openai_api_key
+from danswer.dynamic_configs.interface import ConfigNotFoundError
 from danswer.server.event_loading import router as event_processing_router
 from danswer.server.health import router as health_router
 from danswer.server.manage import router as admin_router
 from danswer.server.search_backend import router as backend_router
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from fastapi import FastAPI
 from fastapi import Request
 from fastapi.exceptions import RequestValidationError
@@ -117,6 +123,17 @@ def get_application() -> FastAPI:
         from danswer.datastores.qdrant.indexing import create_qdrant_collection
         from danswer.configs.app_configs import QDRANT_DEFAULT_COLLECTION
 
+        auth_status = "off" if DISABLE_AUTH else "on"
+        logger.info(f"User auth is turned {auth_status}")
+
+        if not ENABLE_OAUTH:
+            logger.debug("OAuth is turned off")
+        else:
+            if not GOOGLE_OAUTH_CLIENT_ID or not GOOGLE_OAUTH_CLIENT_SECRET:
+                logger.warning("OAuth is turned on but incorrectly configured")
+            else:
+                logger.debug("OAuth is turned on")
+
         logger.info("Warming up local NLP models.")
         warm_up_models()
 

backend/danswer/search/danswer_helper.py

@@ -7,15 +7,22 @@ from danswer.search.search_utils import get_default_intent_model
 from danswer.search.search_utils import get_default_intent_model_tokenizer
 from danswer.search.search_utils import get_default_tokenizer
 from danswer.server.models import HelperResponse
+from danswer.utils.logger import setup_logger
 from danswer.utils.timing import log_function_time
 from transformers import AutoTokenizer  # type:ignore
 
+logger = setup_logger()
+
 
 def count_unk_tokens(text: str, tokenizer: AutoTokenizer) -> int:
     """Unclear if the wordpiece tokenizer used is actually tokenizing anything as the [UNK] token
     It splits up even foreign characters and unicode emojis without using UNK"""
     tokenized_text = tokenizer.tokenize(text)
-    return len([token for token in tokenized_text if token == tokenizer.unk_token])
+    num_unk_tokens = len(
+        [token for token in tokenized_text if token == tokenizer.unk_token]
+    )
+    logger.debug(f"Total of {num_unk_tokens} UNKNOWN tokens found")
+    return num_unk_tokens
 
 
 @log_function_time()
@@ -34,16 +41,26 @@ def query_intent(query: str) -> tuple[SearchType, QueryFlow]:
     if qa > 20:
         # If one class is very certain, choose it still
         if keyword > 70:
-            return SearchType.KEYWORD, QueryFlow.SEARCH
-        if semantic > 70:
-            return SearchType.SEMANTIC, QueryFlow.SEARCH
+            predicted_search = SearchType.KEYWORD
+            predicted_flow = QueryFlow.SEARCH
+        elif semantic > 70:
+            predicted_search = SearchType.SEMANTIC
+            predicted_flow = QueryFlow.SEARCH
         # If it's a QA question, it must be a "Semantic" style statement/question
-        return SearchType.SEMANTIC, QueryFlow.QUESTION_ANSWER
+        else:
+            predicted_search = SearchType.SEMANTIC
+            predicted_flow = QueryFlow.QUESTION_ANSWER
     # If definitely not a QA question, choose between keyword or semantic search
     elif keyword > semantic:
-        return SearchType.KEYWORD, QueryFlow.SEARCH
+        predicted_search = SearchType.KEYWORD
+        predicted_flow = QueryFlow.SEARCH
     else:
-        return SearchType.SEMANTIC, QueryFlow.SEARCH
+        predicted_search = SearchType.SEMANTIC
+        predicted_flow = QueryFlow.SEARCH
+
+    logger.debug(f"Predicted Search: {predicted_search}")
+    logger.debug(f"Predicted Flow: {predicted_flow}")
+    return predicted_search, predicted_flow
 
 
 def recommend_search_flow(

backend/danswer/search/keyword_search.py

@@ -5,7 +5,7 @@ from danswer.chunking.models import InferenceChunk
 from danswer.configs.app_configs import NUM_RETURNED_HITS
 from danswer.datastores.interfaces import IndexFilter
 from danswer.datastores.interfaces import KeywordIndex
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from danswer.utils.timing import log_function_time
 from nltk.corpus import stopwords  # type:ignore
 from nltk.stem import WordNetLemmatizer  # type:ignore
@@ -45,7 +45,7 @@ def retrieve_keyword_documents(
     if not top_chunks:
         filters_log_msg = json.dumps(filters, separators=(",", ":")).replace("\n", "")
         logger.warning(
-            f"Keyword search returned no results...\nfilters: {filters_log_msg}\nedited query: {edited_query}"
+            f"Keyword search returned no results - Filters: {filters_log_msg}\tEdited Query: {edited_query}"
         )
         return None
     return top_chunks

backend/danswer/search/semantic_search.py

@@ -16,7 +16,7 @@ from danswer.search.models import Embedder
 from danswer.search.search_utils import get_default_embedding_model
 from danswer.search.search_utils import get_default_reranking_model_ensemble
 from danswer.server.models import SearchDoc
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from danswer.utils.timing import log_function_time
 from sentence_transformers import SentenceTransformer  # type: ignore
 
@@ -47,12 +47,16 @@ def semantic_reranking(
     chunks: list[InferenceChunk],
 ) -> list[InferenceChunk]:
     cross_encoders = get_default_reranking_model_ensemble()
-    sim_scores = sum([encoder.predict([(query, chunk.content) for chunk in chunks]) for encoder in cross_encoders])  # type: ignore
-    scored_results = list(zip(sim_scores, chunks))
+    sim_scores = [
+        encoder.predict([(query, chunk.content) for chunk in chunks])  # type: ignore
+        for encoder in cross_encoders
+    ]
+    averaged_sim_scores = sum(sim_scores) / len(sim_scores)
+    scored_results = list(zip(averaged_sim_scores, chunks))
     scored_results.sort(key=lambda x: x[0], reverse=True)
     ranked_sim_scores, ranked_chunks = zip(*scored_results)
 
-    logger.debug(f"Reranked similarity scores: {str(ranked_sim_scores)}")
+    logger.debug(f"Reranked similarity scores: {ranked_sim_scores}")
 
     return ranked_chunks
 

backend/danswer/server/event_loading.py

@@ -4,7 +4,7 @@ from danswer.connectors.slack.connector import get_channel_info
 from danswer.connectors.slack.connector import get_thread
 from danswer.connectors.slack.connector import thread_to_doc
 from danswer.utils.indexing_pipeline import build_indexing_pipeline
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from fastapi import APIRouter
 from pydantic import BaseModel
 from pydantic import Extra

backend/danswer/server/manage.py

@@ -57,7 +57,7 @@ from danswer.server.models import RunConnectorRequest
 from danswer.server.models import StatusResponse
 from danswer.server.models import UserByEmail
 from danswer.server.models import UserRoleResponse
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from fastapi import APIRouter
 from fastapi import Depends
 from fastapi import HTTPException

backend/danswer/server/search_backend.py

@@ -21,7 +21,7 @@ from danswer.server.models import HelperResponse
 from danswer.server.models import QAResponse
 from danswer.server.models import QuestionRequest
 from danswer.server.models import SearchResponse
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from fastapi import APIRouter
 from fastapi import Depends
 from fastapi.responses import StreamingResponse
@@ -93,18 +93,23 @@ def direct_qa(
 def stream_direct_qa(
     question: QuestionRequest, user: User = Depends(current_user)
 ) -> StreamingResponse:
+    send_packet_debug_msg = "Sending Packet: {}"
     top_documents_key = "top_documents"
     unranked_top_docs_key = "unranked_top_documents"
     predicted_flow_key = "predicted_flow"
     predicted_search_key = "predicted_search"
 
+    logger.debug(f"Received QA query: {question.query}")
+    logger.debug(f"Query filters: {question.filters}")
+
     def stream_qa_portions() -> Generator[str, None, None]:
+        start_time = time.time()
+
         query = question.query
         collection = question.collection
         filters = question.filters
         use_keyword = question.use_keyword
         offset_count = question.offset if question.offset is not None else 0
-        logger.info(f"Received QA query: {query}")
 
         predicted_search, predicted_flow = query_intent(query)
         if use_keyword is None:
@@ -121,14 +126,15 @@ def stream_direct_qa(
                 query, user_id, filters, QdrantIndex(collection)
             )
         if not ranked_chunks:
-            yield get_json_line(
-                {
-                    top_documents_key: None,
-                    unranked_top_docs_key: None,
-                    predicted_flow_key: predicted_flow,
-                    predicted_search_key: predicted_search,
-                }
-            )
+            logger.debug("No Documents Found")
+            empty_docs_result = {
+                top_documents_key: None,
+                unranked_top_docs_key: None,
+                predicted_flow_key: predicted_flow,
+                predicted_search_key: predicted_search,
+            }
+            logger.debug(send_packet_debug_msg.format(empty_docs_result))
+            yield get_json_line(empty_docs_result)
             return
 
         top_docs = chunks_to_search_docs(ranked_chunks)
@@ -139,6 +145,7 @@ def stream_direct_qa(
             predicted_flow_key: predicted_flow,
             predicted_search_key: predicted_search,
         }
+        logger.debug(send_packet_debug_msg.format(initial_response_dict))
         yield get_json_line(initial_response_dict)
 
         qa_model = get_default_backend_qa_model(timeout=QA_TIMEOUT)
@@ -156,11 +163,12 @@ def stream_direct_qa(
             ):
                 if response_dict is None:
                     continue
-                logger.debug(response_dict)
+                logger.debug(f"Sending packet: {response_dict}")
                 yield get_json_line(response_dict)
         except Exception:
             # exception is logged in the answer_question method, no need to re-log
             pass
+        logger.info(f"Total QA took {time.time() - start_time} seconds")
         return
 
     return StreamingResponse(stream_qa_portions(), media_type="application/json")

backend/danswer/utils/indexing_pipeline.py

@@ -12,7 +12,7 @@ from danswer.datastores.qdrant.store import QdrantIndex
 from danswer.datastores.typesense.store import TypesenseIndex
 from danswer.search.models import Embedder
 from danswer.search.semantic_search import DefaultEmbedder
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 
 logger = setup_logger()
 

backend/danswer/utils/logger.py

@@ -1,8 +1,25 @@
 import logging
 from logging import Logger
 
+from danswer.configs.app_configs import LOG_LEVEL
 
-def setup_logger(name: str = __name__, log_level: int = logging.INFO) -> Logger:
+
+def get_log_level_from_str(log_level_str: str = LOG_LEVEL) -> int:
+    log_level_dict = {
+        "CRITICAL": logging.CRITICAL,
+        "ERROR": logging.ERROR,
+        "WARNING": logging.WARNING,
+        "INFO": logging.INFO,
+        "DEBUG": logging.DEBUG,
+        "NOTSET": logging.NOTSET,
+    }
+
+    return log_level_dict.get(log_level_str.upper(), logging.INFO)
+
+
+def setup_logger(
+    name: str = __name__, log_level: int = get_log_level_from_str()
+) -> Logger:
     logger = logging.getLogger(name)
 
     # If the logger already has handlers, assume it was already configured and return it.

backend/danswer/utils/timing.py

@@ -4,7 +4,7 @@ from typing import Any
 from typing import cast
 from typing import TypeVar
 
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 
 logger = setup_logger()
 

backend/scripts/reset_indexes.py

@@ -3,7 +3,7 @@ from danswer.configs.model_configs import DOC_EMBEDDING_DIM
 from danswer.datastores.typesense.store import create_typesense_collection
 from danswer.utils.clients import get_qdrant_client
 from danswer.utils.clients import get_typesense_client
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from qdrant_client.http.models import Distance
 from qdrant_client.http.models import VectorParams
 from typesense.exceptions import ObjectNotFound  # type: ignore

backend/scripts/save_load_state.py

@@ -22,7 +22,7 @@ from danswer.datastores.qdrant.indexing import list_qdrant_collections
 from danswer.datastores.typesense.store import create_typesense_collection
 from danswer.utils.clients import get_qdrant_client
 from danswer.utils.clients import get_typesense_client
-from danswer.utils.logging import setup_logger
+from danswer.utils.logger import setup_logger
 from qdrant_client.http.models.models import SnapshotDescription
 from typesense.exceptions import ObjectNotFound  # type: ignore