Stub in an initial OpenAI Responses API

Signed-off-by: Ben Browning <bbrownin@redhat.com>

llama_stack/providers/inline/openai_responses/__init__.py

@@ -0,0 +1,19 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from typing import Any, Dict
+
+from llama_stack.apis.datatypes import Api
+
+from .config import OpenAIResponsesImplConfig
+
+
+async def get_provider_impl(config: OpenAIResponsesImplConfig, deps: Dict[Api, Any]):
+    from .openai_responses import OpenAIResponsesImpl
+
+    impl = OpenAIResponsesImpl(config, deps[Api.models], deps[Api.inference])
+    await impl.initialize()
+    return impl

llama_stack/providers/inline/openai_responses/config.py

@@ -0,0 +1,24 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from typing import Any, Dict
+
+from pydantic import BaseModel
+
+from llama_stack.providers.utils.kvstore.config import KVStoreConfig, SqliteKVStoreConfig
+
+
+class OpenAIResponsesImplConfig(BaseModel):
+    kvstore: KVStoreConfig
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> Dict[str, Any]:
+        return {
+            "kvstore": SqliteKVStoreConfig.sample_run_config(
+                __distro_dir__=__distro_dir__,
+                db_name="openai_responses.db",
+            )
+        }

llama_stack/providers/inline/openai_responses/openai_responses.py

@@ -0,0 +1,126 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+import uuid
+from typing import AsyncIterator, List, Optional, cast
+
+from llama_stack.apis.inference.inference import (
+    Inference,
+    OpenAIAssistantMessageParam,
+    OpenAIChatCompletion,
+    OpenAIChatCompletionContentPartTextParam,
+    OpenAIMessageParam,
+    OpenAIUserMessageParam,
+)
+from llama_stack.apis.models.models import Models, ModelType
+from llama_stack.apis.openai_responses import OpenAIResponses
+from llama_stack.apis.openai_responses.openai_responses import (
+    OpenAIResponseObject,
+    OpenAIResponseObjectStream,
+    OpenAIResponseOutputMessage,
+    OpenAIResponseOutputMessageContentOutputText,
+)
+from llama_stack.log import get_logger
+from llama_stack.providers.utils.kvstore import kvstore_impl
+
+from .config import OpenAIResponsesImplConfig
+
+logger = get_logger(name=__name__, category="openai_responses")
+
+OPENAI_RESPONSES_PREFIX = "openai_responses:"
+
+
+class OpenAIResponsesImpl(OpenAIResponses):
+    def __init__(self, config: OpenAIResponsesImplConfig, models_api: Models, inference_api: Inference):
+        self.config = config
+        self.models_api = models_api
+        self.inference_api = inference_api
+
+    async def initialize(self) -> None:
+        self.kvstore = await kvstore_impl(self.config.kvstore)
+
+    async def shutdown(self) -> None:
+        logger.debug("OpenAIResponsesImpl.shutdown")
+        pass
+
+    async def get_openai_response(
+        self,
+        id: str,
+    ) -> OpenAIResponseObject:
+        key = f"{OPENAI_RESPONSES_PREFIX}{id}"
+        response_json = await self.kvstore.get(key=key)
+        if response_json is None:
+            raise ValueError(f"OpenAI response with id '{id}' not found")
+        return OpenAIResponseObject.model_validate_json(response_json)
+
+    async def create_openai_response(
+        self,
+        input: str,
+        model: str,
+        previous_response_id: Optional[str] = None,
+        store: Optional[bool] = True,
+        stream: Optional[bool] = False,
+    ):
+        model_obj = await self.models_api.get_model(model)
+        if model_obj is None:
+            raise ValueError(f"Model '{model}' not found")
+        if model_obj.model_type == ModelType.embedding:
+            raise ValueError(f"Model '{model}' is an embedding model and does not support chat completions")
+
+        messages: List[OpenAIMessageParam] = []
+        if previous_response_id:
+            previous_response = await self.get_openai_response(previous_response_id)
+            messages.append(OpenAIAssistantMessageParam(content=previous_response.output[0].content[0].text))
+
+        messages.append(OpenAIUserMessageParam(content=input))
+
+        chat_response = await self.inference_api.openai_chat_completion(
+            model=model_obj.identifier,
+            messages=messages,
+        )
+        # type cast to appease mypy
+        chat_response = cast(OpenAIChatCompletion, chat_response)
+
+        output_messages = []
+        for choice in chat_response.choices:
+            output_content = ""
+            if isinstance(choice.message.content, str):
+                output_content = choice.message.content
+            elif isinstance(choice.message.content, OpenAIChatCompletionContentPartTextParam):
+                output_content = choice.message.content.text
+            # TODO: handle image content
+            output_messages.append(
+                OpenAIResponseOutputMessage(
+                    id=f"msg_{uuid.uuid4()}",
+                    content=[OpenAIResponseOutputMessageContentOutputText(text=output_content)],
+                    status="completed",
+                )
+            )
+        response = OpenAIResponseObject(
+            created_at=chat_response.created,
+            id=f"resp-{uuid.uuid4()}",
+            model=model_obj.identifier,
+            object="response",
+            status="completed",
+            output=output_messages,
+        )
+
+        if store:
+            # Store in kvstore
+            key = f"{OPENAI_RESPONSES_PREFIX}{response.id}"
+            await self.kvstore.set(
+                key=key,
+                value=response.model_dump_json(),
+            )
+
+        if stream:
+
+            async def async_response() -> AsyncIterator[OpenAIResponseObjectStream]:
+                yield OpenAIResponseObjectStream(response=response)
+
+            return async_response()
+
+        return response

llama_stack/providers/registry/openai_responses.py

@@ -0,0 +1,25 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from typing import List
+
+from llama_stack.providers.datatypes import Api, InlineProviderSpec, ProviderSpec
+
+
+def available_providers() -> List[ProviderSpec]:
+    return [
+        InlineProviderSpec(
+            api=Api.openai_responses,
+            provider_type="inline::openai-responses",
+            pip_packages=[],
+            module="llama_stack.providers.inline.openai_responses",
+            config_class="llama_stack.providers.inline.openai_responses.config.OpenAIResponsesImplConfig",
+            api_dependencies=[
+                Api.models,
+                Api.inference,
+            ],
+        ),
+    ]