Update app.py

app.py

@@ -48,7 +48,7 @@ if not os.path.exists(MODEL_DIR):
 # Load environment variables
 load_dotenv()
 
-
+vector_cache = {}
 ml_models = {}
 
 @asynccontextmanager
@@ -161,13 +161,34 @@ async def run_hackrx(req: RunRequest):
     print(f"chunking done: {end_time}")
     if not chunks:
         return JSONResponse({"error": "No documents could be processed."}, status_code=400)
+
+    doc_url = str(req.documents)  # Assuming it's a URL or unique path
+
+
     start_time2 = time.time()
-    vectorstore = await FAISS.afrom_documents(
-        documents=chunks, 
-        embedding=ml_models["embedder"]
-    )
+    # ✅ Reuse vectorstore if already cached
+    if doc_url in vector_cache:
+        print(f"♻ Using cached vectorstore for: {doc_url}")
+        vectorstore = vector_cache[doc_url]
+    else:
+        print(f"📄 Processing new document: {doc_url}")
+        if not chunks:
+            return JSONResponse({"error": "No documents could be processed."}, status_code=400)
+
+        # Build vectorstore & save to cache
+        vectorstore = await FAISS.afrom_documents(documents=chunks, embedding=ml_models["embedder"])
+        vector_cache[doc_url] = vectorstore  # store in memory cache
+        print(f"✅ Vectorstore cached for: {doc_url}")
     end_time2 = time.time() - start_time2
     print(f"vector done: {end_time2}")
+
+    # start_time2 = time.time()
+    # vectorstore = await FAISS.afrom_documents(
+    #     documents=chunks, 
+    #     embedding=ml_models["embedder"]
+    # )
+    # end_time2 = time.time() - start_time2
+    # print(f"vector done: {end_time2}")
     # dense_retriever = vectorstore.as_retriever(search_type="mmr",search_kwargs={"k": 8})
     dense_retriever = vectorstore.as_retriever(search_type="mmr",search_kwargs={"k": 12 ,"lambda_mult": 0.5} )