feat(code): embed guidelines in the code chat

src/app/api/api_v1/endpoints/code.py

@@ -4,10 +4,11 @@
 # See LICENSE or go to <https://www.apache.org/licenses/LICENSE-2.0> for full license details.
 
 
-from fastapi import APIRouter, Security, status
+from fastapi import APIRouter, Depends, Security, status
 from fastapi.responses import StreamingResponse
 
-from app.api.dependencies import get_token_payload
+from app.api.dependencies import get_guideline_crud, get_token_payload
+from app.crud.crud_guideline import GuidelineCRUD
 from app.models import UserScope
 from app.schemas.code import ChatHistory
 from app.schemas.login import TokenPayload
@@ -20,11 +21,14 @@
 @router.post("/chat", status_code=status.HTTP_200_OK, summary="Chat with our code model")
 async def chat(
     payload: ChatHistory,
+    guidelines: GuidelineCRUD = Depends(get_guideline_crud),
     token_payload: TokenPayload = Security(get_token_payload, scopes=[UserScope.ADMIN, UserScope.USER]),
 ) -> StreamingResponse:
     telemetry_client.capture(token_payload.user_id, event="compute-chat")
+    # Retrieve the guidelines of this user
+    user_guidelines = [g.content for g in await guidelines.fetch_all(filter_pair=("creator_id", token_payload.user_id))]
     # Run analysis
     return StreamingResponse(
-        ollama_client.chat(payload.model_dump()["messages"]).iter_content(chunk_size=8192),
+        ollama_client.chat(payload.model_dump()["messages"], user_guidelines).iter_content(chunk_size=8192),
         media_type="text/event-stream",
     )

src/app/services/ollama.py

@@ -61,6 +61,8 @@ def validate_example_response(response: str) -> Dict[str, str]:
     "(refuse to answer for the rest)."
 )
 
+GUIDELINE_PROMPT = "When answering user requests, you should at all times keep in mind the following software development guidelines:\n"
+
 
 def validate_parsing_response(response: str) -> List[Dict[str, str]]:
     guideline_list = json.loads(response.strip())
@@ -122,15 +124,18 @@ def _chat(
         self,
         system_prompt: str,
         messages: List[Dict[str, str]],
+        guidelines: List[str],
         timeout: int = 20,
     ) -> requests.Response:
+        _guideline_str = "\n-".join(guidelines)
+        _system = system_prompt if len(guidelines) == 0 else f"{system_prompt} {GUIDELINE_PROMPT}-{_guideline_str}"
         return requests.post(
             f"{self.endpoint}/api/chat",
             json={
                 "model": self.model_name,
                 "stream": True,
                 "options": {"temperature": self.temperature},
-                "messages": [{"role": "system", "content": system_prompt}, *messages],
+                "messages": [{"role": "system", "content": _system}, *messages],
                 "keep_alive": "30s",
             },
             stream=True,
@@ -140,9 +145,10 @@ def _chat(
     def chat(
         self,
         messages: List[Dict[str, str]],
+        guidelines: List[str],
         **kwargs,
     ) -> requests.Response:
-        return self._chat(CHAT_PROMPT, messages, **kwargs)
+        return self._chat(CHAT_PROMPT, messages, guidelines, **kwargs)
 
     def parse_guidelines_from_text(
         self,

src/tests/endpoints/test_code.py

@@ -53,7 +53,7 @@
 @pytest.mark.asyncio()
 async def test_chat(
     async_client: AsyncClient,
-    user_session: AsyncSession,
+    guideline_session: AsyncSession,
     user_idx: Union[int, None],
     payload: Dict[str, Any],
     status_code: int,