Update app.py

app.py

@@ -27,9 +27,16 @@ from GoogleNews import GoogleNews
 from bs4 import BeautifulSoup
 import requests
 from urllib.parse import urlparse, urlunparse
-from stock_vector_db import StockVectorDB
+from langchain.embeddings import OpenAIEmbeddings
+from stock_vector_db import *
 from datetime import datetime
 
+
+# Set your HF dataset repo and token
+HF_REPO_ID = "rajat5ranjan/stock-insights"
+HF_TOKEN = st.secrets["hf_token"]  # Store your HF token safely in Streamlit secrets
+
+
 st.set_page_config(layout="wide")
 
 GOOGLE_API_KEY=os.environ['GOOGLE_API_KEY']
@@ -456,23 +463,27 @@ if activities == "Symbol Analysis":
             st.text(res["output_text"])
 elif activities=="News Sentiment":
 
-            # Initialize DB
-    db = StockVectorDB(index_path="vector_index", log_path="vector_log.json")
+
+
+    # Initialize embedding model and vector DB once per session (cache if needed)
+    embedding_model = OpenAIEmbeddings()
+    vector_db = HFVectorDB(hf_repo_id=HF_REPO_ID, hf_token=HF_TOKEN, embedding_model=embedding_model)
+
     
-    log_df = pd.DataFrame(db.log_data)
-    if log_df.empty:
-        st.write("No log")
-    else:
-        log_df["date"] = pd.to_datetime(log_df["date"])
+    # log_df = pd.DataFrame(db.log_data)
+    # if log_df.empty:
+    #     st.write("No log")
+    # else:
+    #     log_df["date"] = pd.to_datetime(log_df["date"])
         
-        with st.sidebar.expander("🕒 History Filters"):
-            selected_date = st.date_input("Pick a date", value=datetime.now().date())
-            filtered = log_df[log_df["date"] == pd.to_datetime(selected_date)]
+    #     with st.sidebar.expander("🕒 History Filters"):
+    #         selected_date = st.date_input("Pick a date", value=datetime.now().date())
+    #         filtered = log_df[log_df["date"] == pd.to_datetime(selected_date)]
         
-            st.write(f"Found {len(filtered)} entries on {selected_date}")
-            for _, row in filtered.iterrows():
-                st.markdown(f"**{row['company']} ({row['ticker']})** — {row['sentiment']} → {row['action']}")
-                st.caption(f"Reason: {row['reason']}")
+    #         st.write(f"Found {len(filtered)} entries on {selected_date}")
+    #         for _, row in filtered.iterrows():
+    #             st.markdown(f"**{row['company']} ({row['ticker']})** — {row['sentiment']} → {row['action']}")
+    #             st.caption(f"Reason: {row['reason']}")
 
 
         
@@ -559,6 +570,16 @@ elif activities=="News Sentiment":
         # Save LLM output to FAISS DB
         today = datetime.now()
         db.store_top_picks(top_picks, today)
+
+        # Add docs to vector DB and save/upload index
+        vector_db.add_documents(docs)
+
+
+        # Save top picks json for backtesting
+        save_top_picks_json(top_picks, today, path="top_picks.jsonl")
+    
+        # Optionally add top picks as documents to vector DB
+        add_top_picks_to_vector_db(vector_db, top_picks, today)
         
         # Layout
         for stock in top_picks: