fix: add model field to chat completion responses (#174)

aidial_adapter_vertexai/chat_completion.py

@@ -51,6 +51,8 @@ async def _get_model(
 
     @dial_exception_decorator
     async def chat_completion(self, request: Request, response: Response):
+        response.set_model(request.deployment_id)
+
         model = await self._get_model(request)
         tools = ToolsConfig.from_request(request)
         static_tools = StaticToolsConfig.from_request(request)

tests/conftest.py

@@ -41,6 +41,8 @@ async def test_http_client() -> AsyncGenerator[httpx.AsyncClient, None]:
         async with httpx.AsyncClient(
             transport=ASGITransport(app),  # type: ignore
             base_url="http://test-app.com",
+            params={"api-version": "dummy-version"},
+            headers={"api-key": "dummy-key"},
         ) as client:
             yield client
 
@@ -51,8 +53,8 @@ def _get_client(deployment_id: str | None = None) -> AsyncAzureOpenAI:
         return AsyncAzureOpenAI(
             azure_endpoint=str(test_http_client.base_url),
             azure_deployment=deployment_id,
-            api_version="",
-            api_key="dummy_key",
+            api_version="dummy-version",
+            api_key="dummy-key",
             max_retries=2,
             timeout=30,
             http_client=test_http_client,

tests/integration_tests/test_chat_completion.py

@@ -207,6 +207,13 @@ def test_case(
             expected=for_all_choices(lambda s: "5" in s),
         )
 
+        test_case(
+            name="model field",
+            messages=[user("test")],
+            max_tokens=1,
+            expected=lambda s: s.response.model == deployment.value,
+        )
+
         test_case(
             name="hello",
             messages=[user('Reply with "Hello"')],

tests/utils/openai.py

@@ -1,6 +1,6 @@
 import json
 import re
-from typing import Any, AsyncGenerator, Callable, List, Optional, TypeVar
+from typing import Any, Callable, List, Optional, TypeVar
 
 import httpx
 from aidial_sdk.chat_completion.request import Attachment, StaticTool
@@ -10,7 +10,10 @@
     TokenizeResponse,
     TokenizeSuccess,
 )
-from aidial_sdk.utils.streaming import merge_chunks
+from aidial_sdk.utils.merge_chunks import (
+    cleanup_indices,
+    merge_chat_completion_chunks,
+)
 from openai import AsyncAzureOpenAI, AsyncStream
 from openai._types import NOT_GIVEN
 from openai.types import CompletionUsage
@@ -215,7 +218,6 @@ async def tokenize(
     tokenize_response = await http_client.post(
         f"openai/deployments/{model_id}/tokenize",
         json=tokenize_request,
-        headers={"api-key": "dummy_key"},
     )
 
     tokenize_response.raise_for_status()
@@ -250,7 +252,7 @@ async def get_response() -> ChatCompletion:
             merged_tools += tools
 
         response = await client.chat.completions.create(
-            model="dummy_model",
+            model="dummy-model",
             messages=messages,
             stream=stream,
             stop=stop,
@@ -267,14 +269,17 @@ async def get_response() -> ChatCompletion:
         )
 
         if isinstance(response, AsyncStream):
+            chunks: List[dict] = []
+            async for chunk in response:
+                chunks.append(chunk.dict())
+
+            response_dict = merge_chat_completion_chunks(*chunks)
 
-            async def generator() -> AsyncGenerator[dict, None]:
-                async for chunk in response:
-                    yield chunk.dict()
+            for choice in response_dict["choices"]:
+                choice["message"] = cleanup_indices(choice["delta"])
+                del choice["delta"]
 
-            response_dict = await merge_chunks(generator())
             response_dict["object"] = "chat.completion"
-            response_dict["model"] = "dummy_model"
 
             return ChatCompletion.parse_obj(response_dict)
         else: