Update app.py

app.py

@@ -1,4 +1,5 @@
 # app.py
+
 import os
 from pathlib import Path
 
@@ -6,10 +7,10 @@ import gradio as gr
 from PIL import Image
 from huggingface_hub import InferenceClient
 
-# ✅ Use community packages to avoid deprecation warnings
+# ✅ Community imports to avoid deprecation
 from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain_community.vectorstores import FAISS
-from langchain_community.llms import HuggingFaceHub
+from langchain_community.llms import HuggingFaceEndpoint
 
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain.chains import RetrievalQA
@@ -24,13 +25,17 @@ FIG_DIR = Path("figures")
 PDF_DIR.mkdir(exist_ok=True)
 FIG_DIR.mkdir(exist_ok=True)
 
+# ————— Read your HF_TOKEN secret —————
+hf_token = os.environ["HF_TOKEN"]
+
 # ————— Embeddings & LLM Setup —————
 embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
 
-# LLM via Hugging Face Inference API
-llm = HuggingFaceHub(
-    repo_id="google/flan-t5-base",
-    model_kwargs={"temperature": 0.5, "max_length": 512}
+# LLM via HF Inference API endpoint
+llm = HuggingFaceEndpoint(
+    endpoint_url="https://api-inference.huggingface.co/models/google/flan-t5-base",
+    huggingfacehub_api_token=hf_token,
+    model_kwargs={"temperature": 0.5, "max_length": 512},
 )
 
 # Prompt
@@ -43,9 +48,9 @@ Answer (up to 3 sentences):
 prompt = PromptTemplate(template=TEMPLATE, input_variables=["context", "question"])
 
 # Inference client for image captioning
-vision_client = InferenceClient("Salesforce/blip-image-captioning-base")
+vision_client = InferenceClient("Salesforce/blip-image-captioning-base", token=hf_token)
 
-# Globals (will set after processing)
+# Globals (will initialize after processing)
 vector_store = None
 qa_chain = None