Update app.py

app.py

@@ -44,9 +44,11 @@ client = InferenceClient(HF_MODEL_ID, token=os.getenv("HF_TOKEN"))
 
 llm = HuggingFaceEndpoint(
     repo_id="ibm-granite/granite-3.1-8b-instruct", # Using Granite for 2026 stability
+    task="text-generation",
     max_new_tokens=512,
     temperature=0.1,
-    huggingfacehub_api_token=os.getenv("HF_TOKEN")
+    huggingfacehub_api_token=os.getenv("HF_TOKEN"),
+    wait_for_model=True
 )
 
 
@@ -122,8 +124,7 @@ def build_retriever(file_path: str):
 # 2. Simplified QA Chain function
 def retriever_qa(file, query):
     retr = build_retriever(file)
-
-    # RetrievalQA works perfectly with HuggingFaceEndpoint
+    
     qa = RetrievalQA.from_chain_type(
         llm=llm,
         chain_type="stuff",
@@ -131,9 +132,9 @@ def retriever_qa(file, query):
         return_source_documents=False,
     )
 
-    # Note: Use the dict format for invoke in 2026 LangChain
-    response = qa.invoke({"query": query})
-    return response["result"]
+    # Use a dictionary for the input
+    result = qa.invoke({"query": query})
+    return result["result"]
 
 
 # -------------------------