feat(chat) : replace chat factory by litellm

pyproject.toml

@@ -7,16 +7,13 @@ readme = "README.md"
 
 [tool.poetry.dependencies]
 python = ">=3.10,<3.13"
-azure-ai-inference = "^1.0.0b8"
 azure-core = "^1.32.0"
 backoff = "^2.2.1"
 ecologits = "^0.5.2"
 fastapi = "^0.115.8"
 huggingface-hub = "^0.29.1"
 langdetect = "^1.0.9"
 litellm = "^1.61.15"
-mistralai = "^1.2.2"
-openai = "^1.60.2"
 psycopg2-binary = "^2.9.10"
 pydantic = "^2.1.1"
 pydantic-settings = "^2.0.3"

rag-metrics.py

@@ -27,7 +27,7 @@
 )
 
 from src.app.api.dependencies import get_settings
-from src.app.services.abst_chat import ChatFactory
+from src.app.services.abst_chat import AbstractChat
 
 logger = logging.getLogger(__name__)
 
@@ -41,9 +41,13 @@
 # init chat client
 settings = get_settings()
 
-chat = ChatFactory().create_chat("openai")
+chat = AbstractChat(
+    model="azure/gpt-4o-mini",
+    API_KEY=settings.AZURE_API_KEY,
+    API_BASE=settings.AZURE_API_BASE,
+    API_VERSION=settings.AZURE_API_VERSION,
+)
 
-chat.init_client()
 
 # init test chat client
 model = "gpt-4o-mini"

src/app/api/api_v1/endpoints/chat.py

@@ -4,10 +4,11 @@
 from fastapi import APIRouter, Depends, HTTPException, status
 from fastapi.responses import StreamingResponse
 from openai import RateLimitError
+from pydantic import BaseModel
 
 from src.app.api.dependencies import get_settings
 from src.app.models import chat as models
-from src.app.services.abst_chat import AbstractChat, ChatFactory
+from src.app.services.abst_chat import AbstractChat
 from src.app.services.constants import subjects as subjectsDict
 from src.app.services.exceptions import (
     EmptyQueryError,
@@ -23,8 +24,12 @@
 
 settings = get_settings()
 
-chatfactory: AbstractChat = ChatFactory().create_chat("openai")
-chatfactory.init_client()
+chatfactory = AbstractChat(
+    model="azure/gpt-4o-mini",
+    API_KEY=settings.AZURE_API_KEY,
+    API_BASE=settings.AZURE_API_BASE,
+    API_VERSION=settings.AZURE_API_VERSION,
+)
 
 
 def get_params(body: models.Context) -> models.ContextOut:
@@ -42,6 +47,10 @@ def get_params(body: models.Context) -> models.ContextOut:
     )
 
 
+class Response(BaseModel):
+    greeting: str
+
+
 @router.post(
     "/reformulate/query",
     summary="Reformulate User Query",

src/app/api/api_v1/endpoints/tutor.py

@@ -4,7 +4,7 @@
 
 from src.app.api.dependencies import get_settings
 from src.app.models.search import EnhancedSearchQuery
-from src.app.services.abst_chat import AbstractChat, ChatFactory
+from src.app.services.abst_chat import AbstractChat
 from src.app.services.exceptions import NoResultsError
 from src.app.services.search import SearchService
 from src.app.services.search_helpers import search_multi_inputs
@@ -23,8 +23,13 @@
 
 settings = get_settings()
 
-chatfactory: AbstractChat = ChatFactory().create_chat("openai", model="gpt-4o")
-chatfactory.init_client()
+
+chatfactory = AbstractChat(
+    model="azure/gpt-4o",
+    API_KEY=settings.AZURE_GPT_4O_API_KEY,
+    API_BASE=settings.AZURE_GPT_4O_API_BASE,
+    API_VERSION=settings.AZURE_GPT_4O_API_VERSION,
+)
 
 sp = SearchService()
 
@@ -69,9 +74,14 @@ async def tutor_search(
     ]
 
     try:
-        themes_extracted = await chatfactory.chat_schema(
-            model="gpt-4o", messages=messages, response_format=ExtractorOuputList  # type: ignore
+        themes_extracted = await chatfactory.chat_client.completion(
+            messages=messages, response_format=ExtractorOuputList
         )
+
+        assert isinstance(themes_extracted, dict)
+
+        themes_extracted = ExtractorOuputList(**themes_extracted)
+
     except Exception as e:
         logger.error(f"Error in chat schema: {e}")
         # todo: handle error