feat(backend): if no temperature is sent with chat request, use agent temperature

Author: ezawadski

src/backend/routers/chat.py

@@ -1,6 +1,6 @@
 from typing import Any, Generator
 
-from fastapi import APIRouter, Depends, Request
+from fastapi import APIRouter, Depends
 from sse_starlette.sse import EventSourceResponse
 
 from backend.chat.custom.custom import CustomChat
@@ -31,7 +31,6 @@
 async def chat_stream(
     session: DBSessionDep,
     chat_request: CohereChatRequest,
-    request: Request,
     ctx: Context = Depends(get_context),
 ) -> Generator[ChatResponseEvent, Any, None]:
     """
@@ -58,7 +57,7 @@ async def chat_stream(
         managed_tools,
         next_message_position,
         ctx,
-    ) = process_chat(session, chat_request, request, ctx)
+    ) = process_chat(session, chat_request, ctx)
 
     return EventSourceResponse(
         generate_chat_stream(
@@ -86,7 +85,6 @@ async def chat_stream(
 async def regenerate_chat_stream(
     session: DBSessionDep,
     chat_request: CohereChatRequest,
-    request: Request,
     ctx: Context = Depends(get_context),
 ) -> EventSourceResponse:
     """
@@ -127,7 +125,7 @@ async def regenerate_chat_stream(
         previous_response_message_ids,
         managed_tools,
         ctx,
-    ) = process_message_regeneration(session, chat_request, request, ctx)
+    ) = process_message_regeneration(session, chat_request, ctx)
 
     return EventSourceResponse(
         generate_chat_stream(
@@ -155,7 +153,6 @@ async def regenerate_chat_stream(
 async def chat(
     session: DBSessionDep,
     chat_request: CohereChatRequest,
-    request: Request,
     ctx: Context = Depends(get_context),
 ) -> NonStreamedChatResponse:
     """
@@ -197,7 +194,7 @@ async def chat(
         managed_tools,
         next_message_position,
         ctx,
-    ) = process_chat(session, chat_request, request, ctx)
+    ) = process_chat(session, chat_request, ctx)
 
     response = await generate_chat_response(
         session,

src/backend/services/chat.py

@@ -4,7 +4,7 @@
 
 import nltk
 from cohere.types import StreamedChatResponse
-from fastapi import HTTPException, Request
+from fastapi import HTTPException
 from fastapi.encoders import jsonable_encoder
 
 from backend.chat.collate import to_dict
@@ -74,19 +74,17 @@ def generate_tools_preamble(chat_request: CohereChatRequest) -> str:
 
 def process_chat(
     session: DBSessionDep,
-    chat_request: BaseChatRequest,
-    request: Request,
+    chat_request: CohereChatRequest,
     ctx: Context,
 ) -> tuple[
-    DBSessionDep, BaseChatRequest, Union[list[str], None], Message, str, str, dict
+    DBSessionDep, CohereChatRequest, Union[list[str], None], Message, str, str, Context
 ]:
     """
     Process a chat request.
 
     Args:
-        chat_request (BaseChatRequest): Chat request data.
+        chat_request (CohereChatRequest): Chat request data.
         session (DBSessionDep): Database session.
-        request (Request): Request object.
         ctx (Context): Context object.
 
     Returns:
@@ -124,6 +122,10 @@ def process_chat(
         chat_request.model = agent.model
         chat_request.preamble = agent.preamble
 
+        # If temperature is not defined in the chat request, use the temperature from the agent
+        if not chat_request.temperature:
+            chat_request.temperature = agent.temperature
+
     should_store = chat_request.chat_history is None and not is_custom_tool_call(
         chat_request
     )
@@ -193,7 +195,6 @@ def process_chat(
 def process_message_regeneration(
     session: DBSessionDep,
     chat_request: CohereChatRequest,
-    request: Request,
     ctx: Context,
 ) -> tuple[Any, CohereChatRequest, Message, list[str], bool, Context]:
     """
@@ -202,7 +203,6 @@ def process_message_regeneration(
     Args:
         session (DBSessionDep): Database session.
         chat_request (CohereChatRequest): Chat request data.
-        request (Request): Request object.
         ctx (Context): Context object.
 
     Returns:
@@ -224,6 +224,10 @@ def process_message_regeneration(
         # Set the agent settings in the chat request
         chat_request.preamble = agent.preamble
 
+        # If temperature is not defined in the chat request, use the temperature from the agent
+        if not chat_request.temperature:
+            chat_request.temperature = agent.temperature
+
     conversation_id = chat_request.conversation_id
     ctx.with_conversation_id(conversation_id)
 

src/backend/tests/unit/factories/agent.py

@@ -19,7 +19,7 @@ class Meta:
     description = factory.Faker("sentence")
     preamble = factory.Faker("sentence")
     version = factory.Faker("random_int")
-    temperature = factory.Faker("pyfloat")
+    temperature = factory.Faker("pyfloat", min_value=0.0, max_value=1.0)
     created_at = factory.Faker("date_time")
     updated_at = factory.Faker("date_time")
     tools = factory.List(

src/backend/tests/unit/routers/test_chat.py

@@ -311,7 +311,6 @@ def test_streaming_fail_chat_missing_message(
                 "loc": ["body", "message"],
                 "msg": "Field required",
                 "input": {},
-                "url": "https://errors.pydantic.dev/2.10/v/missing",
             }
         ]
     }