Update app.py

app.py

@@ -143,24 +143,30 @@ def create_web_search_vectors(search_results):
     
     return FAISS.from_documents(documents, embed)
 
-def summarize_article(article, content, query, model, system_prompt, client, temperature=0.2):
-    prompt = f"""Using the following article:
-    Title: {article.get('title', 'No Title Available')}
-    Content (excerpt): {article.get('body', '')[:2000]}  # Truncate if too long
-    URL: {article.get('href', 'No URL Available')}
-    
-    And based on the following web search context (excerpt):
-    {content[:2000]}  # Truncate if too long
+def summarize_article(article, content, model, system_prompt, user_query, client, temperature=0.2):
+    prompt = f"""Summarize the following article in the context of broader web search results:
+
+Article:
+Title: {article['title']}
+URL: {article['href']}
+Content: {article['body'][:500]}...  # Truncate to avoid extremely long prompts
+
+Additional Context:
+{content[:1000]}...  # Truncate additional context as well
+
+User Query: {user_query}
+
+ Write a detailed and complete research document which addresses the User Query, incorporating both the specific article and the broader context. Focus on the most relevant information.
+"""
+
+    # Calculate input tokens (this is an approximation, you might need a more accurate method)
+    input_tokens = len(prompt.split()) // 4
     
-    Write a detailed and complete research document. The document should include:
-    1. An introduction
-    2. Key findings from both the article and search context
-    3. A conclusion that directly answers the user's request: '{query}'."""
+    # Get the token limit for the current model
+    model_token_limit = MODEL_TOKEN_LIMITS.get(model, 8192)  # Default to 8192 if model not found
     
-    # Calculate token usage and model limits
-    input_tokens = len(prompt.split()) // 4  # Approximate token count
-    model_token_limit = MODEL_TOKEN_LIMITS.get(model, 8192)  # Default limit is 8192 if model is not found
-    max_new_tokens = min(model_token_limit - input_tokens, 6500)  # Cap output tokens to avoid exceeding limits
+    # Calculate max_new_tokens
+    max_new_tokens = min(model_token_limit - input_tokens, 6500)  # Cap at 6500 to be safe
 
     try:
         response = client.chat_completion(