Update app.py

app.py

@@ -6,6 +6,7 @@ import sqlite3
 import json
 import numpy as np
 import datetime
+import re
 from langchain.chains import RetrievalQA
 from langchain.schema import Document
 from langchain_core.retrievers import BaseRetriever
@@ -37,19 +38,23 @@ uploaded_files = st.file_uploader(
     "Upload JSON files in batches (any structure)", type="json", accept_multiple_files=True
 )
 
-# --- Improved Flattening: extracts entity from emails/user fields for better matching
+# --- Enhanced Flattening: extract names from emails/user fields for LLM context
 def flatten_json_obj(obj, parent_key="", sep="."):
     items = {}
     if isinstance(obj, dict):
         for k, v in obj.items():
             new_key = f"{parent_key}{sep}{k}" if parent_key else k
-            # Entity extraction: add name from email
+            # Entity extraction: add name(s) from email/user
             if (
                 k.lower() in {"customer", "user", "email", "username"} and
                 isinstance(v, str) and "@" in v
             ):
                 local = v.split("@")[0]
-                items[new_key + "_name"] = local
+                local_clean = re.sub(r'[^a-zA-Z0-9]', ' ', local)
+                parts = [part for part in local_clean.split() if part]
+                if parts:
+                    items[new_key + "_name"] = parts[0]
+                    items[new_key + "_all_names"] = " ".join(parts)
             items.update(flatten_json_obj(v, new_key, sep=sep))
     elif isinstance(obj, list):
         for i, v in enumerate(obj):
@@ -155,7 +160,6 @@ def python_fuzzy_match(user_query, top_k=5):
     results = []
     for row in cursor.fetchall():
         flat_text = row[4].lower()
-        # score = # of query terms present as substring in the flat_text
         score = sum(any(term in flat_text for term in query_terms) for term in query_terms)
         if score > 0:
             results.append((score, row))
@@ -251,7 +255,6 @@ def send_message():
         return
     st.session_state.messages.append({"role": "user", "content": user_input})
     with st.spinner("Thinking..."):
-        # Correct key: "query"
         result = qa_chain({"query": user_input})
         answer = result['result']
         st.session_state.messages.append({"role": "assistant", "content": answer})