Update src/streamlit_app.py

src/streamlit_app.py

@@ -1,4 +1,4 @@
-# app_hf_space.py
+# app_hf_space.py (Iteration)
 import streamlit as st
 from langchain_huggingface import HuggingFaceEmbeddings
 from langchain_community.vectorstores.faiss import FAISS
@@ -17,16 +17,47 @@ embeddings = HuggingFaceEmbeddings(
     cache_dir=".hf_cache"
 )
 
-# 2. Load FAISS index
+# 2. Preload & Ingest (optional)
+import os, zipfile
+from streamlit import sidebar
+
+# Auto-extract preloaded zip if present
+docs_dir = "docs"
+zip_path = "preloaded_docs.zip"
+if os.path.exists(zip_path):
+    with zipfile.ZipFile(zip_path, "r") as z:
+        z.extractall(docs_dir)
+    sidebar.success(f"Extracted {zip_path} to {docs_dir}/")
+
+# Sidebar button to re-ingest docs and rebuild index
+if sidebar.button("Re-ingest docs & rebuild index"):
+    from ingest import load_documents, text_splitter, embeddings as ingest_embeddings
+    docs = load_documents(docs_dir)
+    chunks = text_splitter.split_documents(docs)
+    FAISS.from_documents(chunks, ingest_embeddings).save_local(INDEX_DIR)
+    sidebar.success("Re-ingestion complete and index rebuilt.")
+
+# 3. Load FAISS index
 store = FAISS.load_local(
     INDEX_DIR,
     embeddings
 )
 
-# 3. Initialize HF LLM via pipeline (inference API)
+# 4. Initialize HF LLM via pipeline (inference API)
+# ...
+# 5. Build RetrievalQA chain
+# ...
+# 6. Streamlit UI
+# ... Initialize HF LLM via pipeline (inference API)
 llm = HuggingFacePipeline.from_model_id(
     model_id=LLM_MODEL_ID,
-    task="text2text-generation"
+    task="text2text-generation",
+    pipeline_kwargs={
+        # Device mapping for inference
+        "device": -1,
+        # Cache directory for model weights
+        "cache_dir": ".hf_cache"
+    }
 )
 
 # 4. Build RetrievalQA chain
@@ -37,13 +68,14 @@ aqa_chain = RetrievalQA.from_chain_type(
 )
 
 # 5. Streamlit UI
-st.title("🦜🔗 RAG App via HF Spaces")
-query = st.text_input("Ask a question about your docs:")
+def main():
+    st.title("🦜🔗 RAG App via HF Spaces")
+    query = st.text_input("Ask a question about your docs:")
 
-if query:
-    with st.spinner("Generating answer via HF Space..."):
-        answer = aqa_chain.run(query)
-    st.markdown(f"**Answer:** {answer}")
+    if query:
+        with st.spinner("Generating answer via HF Space..."):
+            answer = aqa_chain.run(query)
+        st.markdown(f"**Answer:** {answer}")
 
-# NOTE: Deploy this to Hugging Face Spaces for fully-managed hosting.
-# Just push this file to your repo on HF and enable Streamlit space.
+if __name__ == "__main__":
+    main()