Allow empty choices list (#26)

Author: vrslev

any_llm_client/clients/openai.py

@@ -86,7 +86,7 @@ class OneStreamingChoice(pydantic.BaseModel):
 
 
 class ChatCompletionsStreamingEvent(pydantic.BaseModel):
-    choices: typing.Annotated[list[OneStreamingChoice], annotated_types.MinLen(1)]
+    choices: list[OneStreamingChoice]
 
 
 class OneNotStreamingChoiceMessage(pydantic.BaseModel):
@@ -269,7 +269,8 @@ async def _iter_response_chunks(self, response: httpx.Response) -> typing.AsyncI
                 _handle_validation_error(content=event.data.encode(), original_error=validation_error)
 
             if not (
-                (validated_delta := validated_response.choices[0].delta)
+                (validated_choices := validated_response.choices)
+                and (validated_delta := validated_choices[0].delta)
                 and (validated_delta.content or validated_delta.reasoning_content)
             ):
                 continue

tests/test_openai_client.py

@@ -91,6 +91,7 @@ async def test_ok(self, faker: faker.Faker, func_request: LLMFuncRequest) -> Non
                 + ChatCompletionsStreamingEvent(choices=[OneStreamingChoice(delta=one_message)]).model_dump_json()
                 for one_message in generated_messages
             )
+            + f"\n\ndata: {ChatCompletionsStreamingEvent(choices=[]).model_dump_json()}"
             + f"\n\ndata: [DONE]\n\ndata: {faker.pystr()}\n\n"
         )
         response: typing.Final = httpx.Response(
@@ -104,23 +105,6 @@ async def test_ok(self, faker: faker.Faker, func_request: LLMFuncRequest) -> Non
 
         assert result == expected_result
 
-    async def test_fails_without_alternatives(self) -> None:
-        response_content: typing.Final = (
-            f"data: {ChatCompletionsStreamingEvent.model_construct(choices=[]).model_dump_json()}\n\n"
-        )
-        response: typing.Final = httpx.Response(
-            200,
-            headers={"Content-Type": "text/event-stream"},
-            content=response_content,
-        )
-        client: typing.Final = any_llm_client.get_client(
-            OpenAIConfigFactory.build(),
-            transport=httpx.MockTransport(lambda _: response),
-        )
-
-        with pytest.raises(LLMResponseValidationError):
-            await consume_llm_message_chunks(client.stream_llm_message_chunks(**LLMFuncRequestFactory.build()))
-
 
 class TestOpenAILLMErrors:
     @pytest.mark.parametrize("stream", [True, False])