Minor fixed to openai generators.

Author: blkt

src/codegate/providers/openai/provider.py

@@ -18,6 +18,9 @@
 )
 
 
+logger = structlog.get_logger("codegate")
+
+
 class OpenAIProvider(BaseProvider):
     def __init__(
         self,
@@ -74,7 +77,6 @@ async def process_request(
         except Exception as e:
             #  check if we have an status code there
             if hasattr(e, "status_code"):
-                logger = structlog.get_logger("codegate")
                 logger.error("Error in OpenAIProvider completion", error=str(e))
 
                 raise HTTPException(status_code=e.status_code, detail=str(e))  # type: ignore
@@ -106,6 +108,10 @@ async def create_completion(
             req = ChatCompletionRequest.model_validate_json(body)
             is_fim_request = FIMAnalyzer.is_fim_request(request.url.path, req)
 
+            if not req.stream:
+                logger.warn("We got a non-streaming request, forcing to a streaming one")
+                req.stream = True
+
             return await self.process_request(
                 req,
                 api_key,

src/codegate/types/openai/_generators.py

@@ -7,6 +7,7 @@
 import structlog
 
 from ._response_models import (
+    ChatCompletion,
     ErrorDetails,
     MessageError,
     StreamingChatCompletion,
@@ -62,7 +63,6 @@ async def streaming(request, api_key, url):
     payload = request.json(exclude_defaults=True)
     if os.getenv("CODEGATE_DEBUG_OPENAI") is not None:
         print(payload)
-        print(headers)
 
     client = httpx.AsyncClient()
     async with client.stream(
@@ -74,6 +74,11 @@ async def streaming(request, api_key, url):
         # TODO figure out how to best return failures
         match resp.status_code:
             case 200:
+                if not request.stream:
+                    body = await resp.aread()
+                    yield ChatCompletion.model_validate_json(body)
+                    return
+
                 async for message in message_wrapper(resp.aiter_lines()):
                     yield message
             case 400 | 401 | 403 | 404 | 413 | 429:
@@ -115,7 +120,6 @@ async def message_wrapper(lines):
             item = StreamingChatCompletion.model_validate_json(payload)
             yield item
         except Exception as e:
-            print(f"WAAAGH {payload}")
             logger.warn("HTTP error while consuming SSE stream", payload=payload, exc_info=e)
             err = MessageError(
                 error=ErrorDetails(