Update app.py

app.py

@@ -60,18 +60,17 @@ client = InferenceClient(HF_MODEL_ID, token=os.getenv("HF_TOKEN"))
 import os
 from langchain_huggingface import HuggingFaceEndpoint, ChatHuggingFace
 
-# 1. Define the endpoint with text-generation task
-# This is the "engine" that talks to the Hugging Face servers
-http_llm = HuggingFaceEndpoint(
+# 1. Create the endpoint with the 'conversational' task
+llm_endpoint = HuggingFaceEndpoint(
     repo_id="mistralai/Mistral-7B-Instruct-v0.3",
-    task="text-generation",
+    task="conversational",  # Change this from 'text-generation'
     max_new_tokens=512,
     huggingfacehub_api_token=os.getenv("HF_TOKEN"),
 )
 
-# 2. Wrap it manually in ChatHuggingFace
-# We pass it as 'llm' to the constructor
-llm = ChatHuggingFace(llm=http_llm)
+# 2. Wrap it in ChatHuggingFace
+# This is REQUIRED for conversational tasks to work with RAG chains
+llm = ChatHuggingFace(llm=llm_endpoint)
 
 # -------------------------
 # Document loader