Save conversation to DB in the background, as an asyncio task

Author: debanjum

src/khoj/processor/conversation/anthropic/anthropic_chat.py

@@ -1,3 +1,4 @@
+import asyncio
 import logging
 from datetime import datetime, timedelta
 from typing import AsyncGenerator, Dict, List, Optional
@@ -193,13 +194,13 @@ async def converse_anthropic(
     if conversation_commands == [ConversationCommand.Notes] and is_none_or_empty(references):
         response = prompts.no_notes_found.format()
         if completion_func:
-            await completion_func(chat_response=response)
+            asyncio.create_task(completion_func(chat_response=response))
         yield response
         return
     elif conversation_commands == [ConversationCommand.Online] and is_none_or_empty(online_results):
         response = prompts.no_online_results_found.format()
         if completion_func:
-            await completion_func(chat_response=response)
+            asyncio.create_task(completion_func(chat_response=response))
         yield response
         return
 
@@ -251,4 +252,4 @@ async def converse_anthropic(
 
     # Call completion_func once finish streaming and we have the full response
     if completion_func:
-        await completion_func(chat_response=full_response)
+        asyncio.create_task(completion_func(chat_response=full_response))

src/khoj/processor/conversation/google/gemini_chat.py

@@ -1,3 +1,4 @@
+import asyncio
 import logging
 from datetime import datetime, timedelta
 from typing import AsyncGenerator, Dict, List, Optional
@@ -218,13 +219,13 @@ async def converse_gemini(
     if conversation_commands == [ConversationCommand.Notes] and is_none_or_empty(references):
         response = prompts.no_notes_found.format()
         if completion_func:
-            await completion_func(chat_response=response)
+            asyncio.create_task(completion_func(chat_response=response))
         yield response
         return
     elif conversation_commands == [ConversationCommand.Online] and is_none_or_empty(online_results):
         response = prompts.no_online_results_found.format()
         if completion_func:
-            await completion_func(chat_response=response)
+            asyncio.create_task(completion_func(chat_response=response))
         yield response
         return
 
@@ -275,4 +276,4 @@ async def converse_gemini(
 
     # Call completion_func once finish streaming and we have the full response
     if completion_func:
-        await completion_func(chat_response=full_response)
+        asyncio.create_task(completion_func(chat_response=full_response))

src/khoj/processor/conversation/offline/chat_model.py

@@ -202,13 +202,13 @@ async def converse_offline(
     if conversation_commands == [ConversationCommand.Notes] and is_none_or_empty(references):
         response = prompts.no_notes_found.format()
         if completion_func:
-            await completion_func(chat_response=response)
+            asyncio.create_task(completion_func(chat_response=response))
         yield response
         return
     elif conversation_commands == [ConversationCommand.Online] and is_none_or_empty(online_results):
         response = prompts.no_online_results_found.format()
         if completion_func:
-            await completion_func(chat_response=response)
+            asyncio.create_task(completion_func(chat_response=response))
         yield response
         return
 
@@ -317,7 +317,7 @@ def _sync_llm_thread():
 
     # Call the completion function after streaming is done
     if completion_func:
-        await completion_func(chat_response=aggregated_response_container["response"])
+        asyncio.create_task(completion_func(chat_response=aggregated_response_container["response"]))
 
 
 def send_message_to_model_offline(

src/khoj/processor/conversation/openai/gpt.py

@@ -1,3 +1,4 @@
+import asyncio
 import logging
 from datetime import datetime, timedelta
 from typing import AsyncGenerator, Dict, List, Optional
@@ -219,13 +220,13 @@ async def converse_openai(
     if conversation_commands == [ConversationCommand.Notes] and is_none_or_empty(references):
         response = prompts.no_notes_found.format()
         if completion_func:
-            await completion_func(chat_response=response)
+            asyncio.create_task(completion_func(chat_response=response))
         yield response
         return
     elif conversation_commands == [ConversationCommand.Online] and is_none_or_empty(online_results):
         response = prompts.no_online_results_found.format()
         if completion_func:
-            await completion_func(chat_response=response)
+            asyncio.create_task(completion_func(chat_response=response))
         yield response
         return
 
@@ -277,7 +278,7 @@ async def converse_openai(
 
     # Call completion_func once finish streaming and we have the full response
     if completion_func:
-        await completion_func(chat_response=full_response)
+        asyncio.create_task(completion_func(chat_response=full_response))
 
 
 def clean_response_schema(schema: BaseModel | dict) -> dict:

src/khoj/routers/api_chat.py

@@ -998,22 +998,26 @@ def collect_telemetry():
                 return
 
             llm_response = construct_automation_created_message(automation, crontime, query_to_run, subject)
-            await save_to_conversation_log(
-                q,
-                llm_response,
-                user,
-                meta_log,
-                user_message_time,
-                intent_type="automation",
-                client_application=request.user.client_app,
-                conversation_id=conversation_id,
-                inferred_queries=[query_to_run],
-                automation_id=automation.id,
-                query_images=uploaded_images,
-                train_of_thought=train_of_thought,
-                raw_query_files=raw_query_files,
-                tracer=tracer,
+            # Trigger task to save conversation to DB
+            asyncio.create_task(
+                save_to_conversation_log(
+                    q,
+                    llm_response,
+                    user,
+                    meta_log,
+                    user_message_time,
+                    intent_type="automation",
+                    client_application=request.user.client_app,
+                    conversation_id=conversation_id,
+                    inferred_queries=[query_to_run],
+                    automation_id=automation.id,
+                    query_images=uploaded_images,
+                    train_of_thought=train_of_thought,
+                    raw_query_files=raw_query_files,
+                    tracer=tracer,
+                )
             )
+            # Send LLM Response
             async for result in send_llm_response(llm_response, tracer.get("usage")):
                 yield result
             return