Update app.py

app.py

@@ -419,12 +419,31 @@ After writing the document, please provide a list of sources used in your respon
     if not full_response:
         yield "I apologize, but I couldn't generate a response at this time. Please try again later."
 
+def create_web_search_vectors(search_results):
+    embed = get_embeddings()
+    
+    documents = []
+    for result in search_results:
+        if 'body' in result:
+            content = f"{result['title']}\n{result['body']}\nSource: {result['href']}"
+            documents.append(Document(page_content=content, metadata={"source": result['href']}))
+    
+    return FAISS.from_documents(documents, embed)
+
 def get_response_with_search(query, model, num_calls=3, temperature=0.2):
     search_results = duckduckgo_search(query)
-    context = "\n".join(f"{result['title']}\n{result['body']}\nSource: {result['href']}\n" 
-                        for result in search_results if 'body' in result)
+    web_search_database = create_web_search_vectors(search_results)
+    
+    if not web_search_database:
+        yield "No web search results available. Please try again.", ""
+        return
+    
+    retriever = web_search_database.as_retriever(search_kwargs={"k": 5})
+    relevant_docs = retriever.get_relevant_documents(query)
+    
+    context = "\n".join([doc.page_content for doc in relevant_docs])
     
-    prompt = f"""Using the following context:
+    prompt = f"""Using the following context from web search results:
 {context}
 You are an expert AI assistant, write a detailed and complete research document that fulfills the following user request: '{query}'
 Base your entire response strictly on the information retrieved from trusted sources. Importantly, only include information that is directly supported by the retrieved content.