Update app.py

app.py

@@ -53,16 +53,21 @@ client = InferenceClient(HF_MODEL_ID, token=os.getenv("HF_TOKEN"))
 #     wait_for_model=True 
 # )
 
-# 1. First, define the base endpoint (keep task as text-generation here)
+# 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", 
+    task="text-generation", # Keep this as text-generation
     max_new_tokens=512,
     huggingfacehub_api_token=os.getenv("HF_TOKEN"),
 )
 
-# 2. Wrap it in ChatHuggingFace to fix the task mismatch
-# This automatically handles the "conversational" formatting for you
+# 2. IMPORTANT: Manually set the model_id before wrapping
+# This bypasses the automated router check that's causing your ValueError
+llm_endpoint.model_id = "mistralai/Mistral-7B-Instruct-v0.3"
+
+# 3. Wrap in ChatHuggingFace
+# This will now handle the chat formatting without the router rejecting the task
 llm = ChatHuggingFace(llm=llm_endpoint)
 
 # -------------------------