Load index

src/rag.py

@@ -1,52 +1,59 @@
+# src/rag.py
 import os
 import pickle
 import faiss
+import numpy as np
 from sentence_transformers import SentenceTransformer
 from groq import Groq
+from ingest import build_index  # 👈 important
 
 INDEX_DIR = "src/index"
-INDEX_FILE = f"{INDEX_DIR}/faiss.index"
-DOC_FILE = f"{INDEX_DIR}/documents.pkl"
+TOP_K = 4
+
+embedder = SentenceTransformer("all-MiniLM-L6-v2")
 
 GROQ_API_KEY = os.environ.get("GROQ_API_KEY")
 if not GROQ_API_KEY:
-    raise RuntimeError("❌ GROQ_API_KEY not set in HF Secrets")
+    raise RuntimeError("❌ GROQ_API_KEY not set in Hugging Face Secrets")
 
 client = Groq(api_key=GROQ_API_KEY)
-embedder = SentenceTransformer("all-MiniLM-L6-v2")
-
-_index = None
-_documents = None
 
 
 def load_index():
-    global _index, _documents
+    index_path = f"{INDEX_DIR}/faiss.index"
+    docs_path = f"{INDEX_DIR}/documents.pkl"
+
+    # 🔥 AUTO INGEST IF MISSING
+    if not os.path.exists(index_path) or not os.path.exists(docs_path):
+        print("⚠️ FAISS index missing. Running ingestion...")
+        build_index()
 
-    if _index is None or _documents is None:
-        if not os.path.exists(INDEX_FILE):
-            raise RuntimeError(
-                "❌ FAISS index not found. Ingestion must run first."
-            )
+    index = faiss.read_index(index_path)
 
-        _index = faiss.read_index(INDEX_FILE)
-        with open(DOC_FILE, "rb") as f:
-            _documents = pickle.load(f)
+    with open(docs_path, "rb") as f:
+        documents = pickle.load(f)
 
-    return _index, _documents
+    return index, documents
 
 
-def retrieve(query, top_k=4):
+def retrieve(query, top_k=TOP_K):
     index, documents = load_index()
 
-    q_emb = embedder.encode([query]).astype("float32")
-    _, indices = index.search(q_emb, top_k)
+    query_embedding = embedder.encode([query]).astype("float32")
+    distances, indices = index.search(query_embedding, top_k)
+
+    results = []
+    for idx in indices[0]:
+        if idx == -1:
+            continue
+        results.append(documents[idx])
 
-    return [documents[i] for i in indices[0] if i != -1]
+    return results
 
 
 def ask_llm(query, contexts):
     context_text = "\n\n".join(
-        f"[{c['source']} p.{c['page']}]\n{c['text']}"
+        f"[p.{c['page']}]\n{c['text']}"
         for c in contexts
     )
 

src/streamlit_app.py

@@ -1,20 +1,11 @@
-import os
+# src/streamlit_app.py
 import streamlit as st
+from rag import retrieve, ask_llm
 
-INDEX_FILE = "src/index/faiss.index"
+st.set_page_config(page_title="OTT Astrology Bot", layout="wide")
+st.title("🪐 OTT Astrology Assistant")
 
-st.set_page_config(page_title="Astrology RAG Bot", layout="wide")
-st.title("🔮 Astrology Knowledge Bot")
-
-# 🚀 Run ingestion ONCE if index is missing
-if not os.path.exists(INDEX_FILE):
-    with st.spinner("📚 Preparing knowledge base (first run only)..."):
-        import ingest  # runs ingestion
-    st.success("✅ Knowledge base ready!")
-
-from rag import retrieve, ask_llm  # SAFE to import now
-
-query = st.text_input("Ask your astrology question")
+query = st.text_input("Ask your astrology question:")
 
 if query:
     with st.spinner("Thinking..."):
@@ -23,7 +14,3 @@ if query:
 
     st.markdown("### 🪐 Answer")
     st.write(answer)
-
-    st.markdown("### 📘 References")
-    for c in contexts:
-        st.write(f"{c['source']} — page {c['page']}")