chore: switch to agentic llm

Author: guenhter

backend/config.yaml

@@ -10,7 +10,7 @@ modules:
       chat_openai: chat_openai
       session: "session"
   chat_openai:
-    class: modai.modules.chat.openai_raw_chat.OpenAILLMChatModule
+    class: modai.modules.chat.openai_agent_chat.StrandsAgentChatModule
     module_dependencies:
       llm_provider_module: openai_model_provider
   model_provider_store:

backend/src/modai/default_config.yaml

@@ -9,7 +9,7 @@ modules:
     module_dependencies:
       chat_openai: chat_openai
   chat_openai:
-    class: modai.modules.chat.openai_raw_chat.OpenAILLMChatModule
+    class: modai.modules.chat.openai_agent_chat.StrandsAgentChatModule
     module_dependencies:
       llm_provider_module: openai_model_provider
   model_provider_store:

…odai/modules/chat/__tests__/test_chat.py …s/chat/__tests__/test_openai_raw_chat.pybackend/src/modai/modules/chat/__tests__/test_chat.py renamed to backend/src/modai/modules/chat/__tests__/test_openai_raw_chat.py backend/src/modai/modules/chat/__tests__/test_chat.py renamed to backend/src/modai/modules/chat/__tests__/test_openai_raw_chat.py

@@ -8,7 +8,13 @@
 from fastapi.responses import StreamingResponse
 from typing import Any, AsyncGenerator
 from modai.module import ModaiModule, ModuleDependencies
-import openai
+from openai.types.responses import (
+    Response as OpenAIResponse,
+    ResponseStreamEvent as OpenAIResponseStreamEvent,
+)
+from openai.types.responses.response_create_params import (
+    ResponseCreateParams as OpenAICreateResponse,
+)
 
 
 class ChatWebModule(ModaiModule, ABC):
@@ -33,8 +39,8 @@ def __init__(self, dependencies: ModuleDependencies, config: dict[str, Any]):
     async def responses_endpoint(
         self,
         request: Request,
-        body_json: openai.types.responses.ResponseCreateParams = Body(...),
-    ) -> openai.types.responses.Response | StreamingResponse:
+        body_json: OpenAICreateResponse = Body(...),
+    ) -> OpenAIResponse | StreamingResponse:
         """
         Handles responses requests. Must be implemented by concrete implementations.
         Fully OpenAI /responses API compatible.
@@ -61,11 +67,8 @@ def __init__(self, dependencies: ModuleDependencies, config: dict[str, Any]):
 
     @abstractmethod
     async def generate_response(
-        self, request: Request, body_json: openai.types.responses.ResponseCreateParams
-    ) -> (
-        openai.types.responses.Response
-        | AsyncGenerator[openai.types.responses.ResponseStreamEvent, None]
-    ):
+        self, request: Request, body_json: OpenAICreateResponse
+    ) -> OpenAIResponse | AsyncGenerator[OpenAIResponseStreamEvent, None]:
         """
         Generate a streaming or non-streaming chat response.
 

backend/src/modai/modules/chat/module.py

@@ -14,6 +14,8 @@
 )
 
 
+# The module is at the moment unuses as I'm unsure if we should support a raw LLM Access as the agentic
+# chat is more what a chat backend needs
 class OpenAILLMChatModule(ChatLLMModule):
     """
     OpenAI LLM Provider for Chat Responses.