fix(openai): user key now used for llm model

backend/core/llm/base.py

@@ -152,7 +152,8 @@ def generate_answer(self, question: str) -> str:
             This function should also call: _create_qa, get_chat_history and format_chat_history.
             It should also update the chat_history in the DB.
             """
-
+
+
         async def generate_stream(self, question: str) -> AsyncIterable:
             """
             Generate a streaming answer to a given question using QA Chain.

backend/core/llm/openai.py

@@ -46,7 +46,7 @@ def embeddings(self) -> OpenAIEmbeddings:
             openai_api_key=self.openai_api_key
         )  # pyright: ignore reportPrivateUsage=none
 
-    def _create_llm(self, model, streaming=False, callbacks=None) -> BaseLLM:
+    def _create_llm(self, model, temperature=0, streaming=False, callbacks=None) -> BaseLLM:
         """
         Determine the language model to be used.
         :param model: Language model name to be used.
@@ -55,9 +55,10 @@ def _create_llm(self, model, streaming=False, callbacks=None) -> BaseLLM:
         :return: Language model instance
         """
         return ChatOpenAI(
-            temperature=self.temperature,
+            temperature=temperature,
             model=model,
             streaming=streaming,
             verbose=True,
             callbacks=callbacks,
+            openai_api_key=self.openai_api_key,
         )  # pyright: ignore reportPrivateUsage=none

backend/core/llm/qa_base.py

@@ -13,7 +13,6 @@
 from repository.chat.update_chat_history import update_chat_history
 from supabase.client import Client, create_client
 from vectorstore.supabase import CustomSupabaseVectorStore
-from langchain.chat_models import ChatOpenAI
 from repository.chat.update_message_by_id import update_message_by_id
 import json
 
@@ -175,12 +174,8 @@ async def generate_stream(self, question: str) -> AsyncIterable:
         callback = self.callbacks[0]
         callback = AsyncIteratorCallbackHandler()
         self.callbacks = [callback]
-        model = ChatOpenAI(
-            streaming=True,
-            verbose=True,
-            callbacks=[callback],
-        )
-        llm = ChatOpenAI(temperature=0)
+        model = self._create_llm(model=self.model, streaming=True, callbacks=self.callbacks)
+        llm = self._create_llm(model=self.model,temperature=self.temperature)
         question_generator = LLMChain(llm=llm, prompt=CONDENSE_QUESTION_PROMPT)
         doc_chain = load_qa_chain(model, chain_type="stuff")
         qa = ConversationalRetrievalChain(

frontend/lib/config/CONSTANTS.ts

@@ -1,4 +1,4 @@
 export const GITHUB_URL = "https://github.com/stangirard/quivr";
 export const TWITTER_URL = "https://twitter.com/quivr_brain";
 export const DISCORD_URL = "https://discord.gg/HUpRgp2HG8";
-export const DEFAULT_BACKEND_URL = "http://localhost:5000";
+export const DEFAULT_BACKEND_URL = "http://localhost:5050";