Update app.py

app.py

@@ -55,11 +55,17 @@ client = InferenceClient(HF_MODEL_ID, token=os.getenv("HF_TOKEN"))
 
 # 1. Configure the endpoint
 # We use 'text-generation' because that's what the LangChain client expects
-llm_endpoint = HuggingFaceEndpoint(
-    repo_id="mistralai/Mistral-7B-Instruct-v0.3",
-    task="text-generation", # Keep this as text-generation
-    max_new_tokens=512,
-    huggingfacehub_api_token=os.getenv("HF_TOKEN"),
+# Use the dedicated Chat constructor
+# This handles the model_id and the conversational task mapping automatically
+llm = ChatHuggingFace.from_model_id(
+    model_id="mistralai/Mistral-7B-Instruct-v0.3",
+    model_kwargs={
+        "max_new_tokens": 512,
+        "huggingfacehub_api_token": os.getenv("HF_TOKEN"),
+    },
+    # This tells the server we are sending a Chat request
+    # but the local wrapper knows it's an "Instruct" model.
+    task="text-generation", 
 )
 
 # 2. IMPORTANT: Manually set the model_id before wrapping