Update app.py

app.py

@@ -74,7 +74,7 @@ class RateLimiter:
         self.max_calls = max_calls
         self.time_window = time_window
         self.calls = []
-    
+
     def is_allowed(self):
         now = time.time()
         self.calls = [call for call in self.calls if now - call < self.time_window]
@@ -90,30 +90,7 @@ feedback_data = []
 
 def get_preloaded_context():
     """Get preloaded context information"""
-    context = f"""{FORMATTED_DATE_TIME}
-System Information: You are an AI assistant with access to current information through web search and academic research tools.
-Always provide sources for factual information.
-
-Available APIs: 
-- Web Search (Tavily)
-- Academic Research (arXiv, Semantic Scholar)
-- PDF Document Analysis
-- Web Page Content Extraction
-
-Specialized Features:
-- Research-focused queries automatically processed
-- Academic paper analysis and summarization
-- Literature review generation
-- Citation management and bibliography creation
-
-Response Guidelines:
-1. After completing your analysis, ALWAYS end with either:
-   '[ANALYSIS COMPLETE]' - when you've fully addressed the query
-   '[FURTHER RESEARCH NEEDED]' - when additional investigation would be beneficial
-
-2. For search results, provide clear synthesis rather than just listing findings
-3. Include specific citations and sources where applicable
-4. Structure complex answers with clear sections when appropriate"""
+    context = f"""{FORMATTED_DATE_TIME} System Information: You are an AI assistant with access to current information through web search and academic research tools. Always provide sources for factual information. Available APIs:  - Web Search (Tavily) - Academic Research (arXiv, Semantic Scholar) - PDF Document Analysis - Web Page Content Extraction Specialized Features: - Research-focused queries automatically processed - Academic paper analysis and summarization - Literature review generation - Citation management and bibliography creation Response Guidelines: 1. After completing your analysis, ALWAYS end with either:    '[ANALYSIS COMPLETE]' - when you've fully addressed the query    '[FURTHER RESEARCH NEEDED]' - when additional investigation would be beneficial 2. For search results, provide clear synthesis rather than just listing findings 3. Include specific citations and sources where applicable 4. Structure complex answers with clear sections when appropriate"""
     return context
 
 def clean_query_for_current_info(query):
@@ -128,8 +105,8 @@ def clean_query_for_current_info(query):
 def determine_research_content_type(query):
     """Determine if query requires research-focused search"""
     research_keywords = [
-        'research', 'study', 'paper', 'academic', 'scientific', 
-        'experiment', 'findings', 'discovery', 'theory', 
+        'research', 'study', 'paper', 'academic', 'scientific',
+        'experiment', 'findings', 'discovery', 'theory',
         'hypothesis', 'methodology', 'conclusion', 'literature',
         'peer reviewed', 'scholarly', 'journal', 'publication',
         'analyze', 'investigate', 'examine', 'review'
@@ -226,7 +203,7 @@ def truncate_history(messages, max_tokens=4000):
         
         if current_tokens + message_tokens > max_tokens:
             break
-            
+        
         truncated.insert(0, message)
         current_tokens += message_tokens
     
@@ -241,10 +218,16 @@ def manage_conversation_memory(messages, max_turns=10):
         return system_msg + recent_messages if system_msg else recent_messages
     return messages
 
-     # Clean query for current info
-     clean_query = clean_query_for_current_info(query)
+def tavily_search(query):
+    """Perform search using Tavily"""
+    if not TAVILY_AVAILABLE or not tavily_client:
+        return "Web search not available."
+    
+    try:
+        # Clean query for current info
+        clean_query = clean_query_for_current_info(query)
         
-    if not clean_query:
+        if not clean_query:
             return "No valid search query provided."
         
         response = tavily_client.search(
@@ -295,7 +278,7 @@ def download_and_extract_pdf(url):
         
         full_text = "\n".join(text_content)
         return f"PDF CONTENT EXTRACTED FROM {url}:\n{full_text[:4000]}..."  # Limit size
-        
+    
     except Exception as e:
         return f"PDF extraction error: {str(e)}"
 
@@ -329,7 +312,7 @@ def scrape_web_page(url):
         content = ' '.join([p.get_text().strip() for p in paragraphs[:30] if p.get_text().strip()])
         
         return f"WEB PAGE CONTENT FROM {url}:\nTitle: {title}\nContent: {content[:3000]}..."  # Limit content size
-        
+    
     except Exception as e:
         return f"Error scraping page: {str(e)}"
 
@@ -360,7 +343,7 @@ def arxiv_search(query):
             return "\n\n---\n\n".join(results)
         else:
             return "No arXiv papers found for this query."
-        
+    
     except Exception as e:
         return f"arXiv search error: {str(e)}"
 
@@ -400,7 +383,7 @@ def semantic_scholar_search(query):
             return "\n\n---\n\n".join(results)
         else:
             return "No Semantic Scholar papers found for this query."
-        
+    
     except Exception as e:
         return f"Semantic Scholar search error: {str(e)}"
 
@@ -412,7 +395,7 @@ def comprehensive_research(query):
     results.append(f"COMPREHENSIVE RESEARCH RESULTS FOR: '{query}'\n" + "="*50)
     
     # Academic databases
-    if TAVILY_AVAILABLE:
+    if TAVILY_AVAILABLE and tavily_client:
         tavily_result = tavily_search(query)
         results.append(f"TAVILY ACADEMIC SEARCH RESULTS:\n{tavily_result}")
     
@@ -433,25 +416,10 @@ def comprehensive_research(query):
     
     return "\n\n---\n\n".join(results)
 
-def perform_search(query):
-    """Perform search using Tavily"""
-    if TAVILY_AVAILABLE and tavily_client:
-        web_result = tavily_search(query)
-        return f"[SEARCH RESULTS FOR '{query}']:\nSource: Web Search\n{web_result}"
-    else:
-        return "Web search not available."
 def analyze_search_results(query, search_results):
     """Create a prompt for the model to analyze search results"""
-    analysis_prompt = f"""Based on the search results below, please answer the original question: "{query}"
-
-Search Results: {search_results}
-
-Please provide a clear, concise answer based on these sources. Include specific names, facts, and cite the sources where possible. Do not mention that you are analyzing search results - just provide the answer directly.
-
-Structure your response thoughtfully and when you complete your analysis, please explicitly state '[ANALYSIS COMPLETE]' at the end if you have fully addressed the query and have no further input. 
-
-If additional research or clarification would be beneficial, please state '[FURTHER RESEARCH NEEDED]'."""
-
+    analysis_prompt = f"""Based on the search results below, please answer the original question: "{query}" Search Results: {search_results} Please provide a clear, concise answer based on these sources. Include specific names, facts, and cite the sources where possible. Do not mention that you are analyzing search results - just provide the answer directly. Structure your response thoughtfully and when you complete your analysis, please explicitly state '[ANALYSIS COMPLETE]' at the end if you have fully addressed the query and have no further input.  If additional research or clarification would be beneficial, please state '[FURTHER RESEARCH NEEDED]'."""
+    
     return analysis_prompt
 
 def generate_bibliography(search_results):
@@ -491,21 +459,7 @@ def generate_bibliography(search_results):
 
 def generate_literature_review(topic, search_results):
     """Generate structured literature review from search results"""
-    prompt = f"""Based on the following research on '{topic}', create a structured literature review:
-
-{search_results}
-
-Please organize your response as follows:
-1. INTRODUCTION: Brief overview of the topic
-2. KEY FINDINGS: Major discoveries and insights from the research
-3. METHODOLOGIES: Common research approaches used
-4. LIMITATIONS: Identified gaps or limitations in current research
-5. FUTURE DIRECTIONS: Suggested areas for future investigation
-6. CONCLUSION: Summary of the current state of research
-
-Format your response clearly with these section headings.
-
-When you complete your analysis, please explicitly state '[ANALYSIS COMPLETE]' at the end."""
+    prompt = f"""Based on the following research on '{topic}', create a structured literature review: {search_results} Please organize your response as follows: 1. INTRODUCTION: Brief overview of the topic 2. KEY FINDINGS: Major discoveries and insights from the research 3. METHODOLOGIES: Common research approaches used 4. LIMITATIONS: Identified gaps or limitations in current research 5. FUTURE DIRECTIONS: Suggested areas for future investigation 6. CONCLUSION: Summary of the current state of research Format your response clearly with these section headings. When you complete your analysis, please explicitly state '[ANALYSIS COMPLETE]' at the end."""
     
     return prompt
 
@@ -548,7 +502,7 @@ def check_analysis_status(content):
     
     # Check if this appears to be a final answer/response
     elif any(phrase in content.lower() for phrase in [
-        "in conclusion", "to summarize", "in summary", 
+        "in conclusion", "to summarize", "in summary",
         "overall", "therefore", "thus", "in closing"
     ]):
         return "✅ AI appears to be concluding its response."
@@ -563,7 +517,7 @@ def check_analysis_status(content):
     
     # Check for literature review generation
     elif "LITERATURE REVIEW" in content and any(header in content for header in [
-        "INTRODUCTION", "KEY FINDINGS", "METHODOLOGIES", 
+        "INTRODUCTION", "KEY FINDINGS", "METHODOLOGIES",
         "LIMITATIONS", "FUTURE DIRECTIONS", "CONCLUSION"
     ]):
         return "📑 Literature review structured - comprehensive analysis provided."
@@ -627,13 +581,13 @@ def generate_with_streaming(messages, model, max_tokens=8192, temperature=0.7, t
                                 continue
         else:
             yield f"Error: {response.status_code} - {response.text}"
-            
+    
     except Exception as e:
         yield f"Connection error: {str(e)}"
     finally:
         end_time = time.time()
         # Track usage (simplified)
-        track_usage("user123", str(messages[-1]) if messages else "", 
+        track_usage("user123", str(messages[-1]) if messages else "",
                    end_time - start_time, len(str(messages)))
 
 def format_code_blocks(text):
@@ -684,7 +638,7 @@ def collect_feedback(feedback, query, response):
 @lru_cache(maxsize=100)
 def cached_search(query):
     """Cache frequent searches"""
-    return perform_search(query)
+    return tavily_search(query)
 
 def handle_api_failure(error_type, fallback_strategy="retry"):
     """Handle different types of API failures gracefully"""
@@ -695,7 +649,7 @@ def export_conversation(chat_history, export_format):
     """Export conversation in various formats"""
     if not chat_history:
         return "No conversation to export"
-        
+    
     if export_format == "JSON":
         # Filter out system messages for export
         exportable_history = [msg for msg in chat_history if msg[0] != "system"]
@@ -731,8 +685,8 @@ def process_url_content(url):
     else:
         return scrape_web_page(url)
 
-def respond(message, chat_history, model_choice, max_tokens, temperature, top_p, 
-           creativity, precision, system_prompt, use_web_search, research_mode, theme):
+def respond(message, chat_history, model_choice, max_tokens, temperature, top_p,
+            creativity, precision, system_prompt, use_web_search, research_mode, theme):
     """Main response handler with conversation history"""
     if not message:
         yield "", chat_history, "", gr.update(choices=[], visible=False), "", "💬 Ready for your query"
@@ -808,7 +762,7 @@ def respond(message, chat_history, model_choice, max_tokens, temperature, top_p,
             bibliography = generate_bibliography(search_result)
             analysis_status = "📚 Comprehensive research completed"
         else:
-            search_result = perform_search(message)
+            search_result = tavily_search(message)
             bibliography = ""
             analysis_status = "📊 Search results retrieved"
         
@@ -859,7 +813,7 @@ def respond(message, chat_history, model_choice, max_tokens, temperature, top_p,
             # Break infinite loops
             if is_looping_content(full_response):
                 # Force search instead of looping
-                search_result = perform_search(message)
+                search_result = tavily_search(message)
                 follow_ups = generate_follow_up_questions(search_result)
                 analysis_status = "⚠️ Loop detected - performing search instead"
                 # Convert back to Gradio format
@@ -875,7 +829,7 @@ def respond(message, chat_history, model_choice, max_tokens, temperature, top_p,
     # Check for tool calls after completion or break loops
     if is_looping_content(full_response):
         # Force search for looping content
-        search_result = perform_search(message)
+        search_result = tavily_search(message)
         follow_ups = generate_follow_up_questions(search_result)
         analysis_status = "⚠️ Loop detected - performing search instead"
         # Convert back to Gradio format
@@ -949,14 +903,14 @@ with gr.Blocks(title="GPT-OSS Research Assistant") as demo:
             clear = gr.Button("Clear")
             theme_toggle = gr.Radio(choices=["Light", "Dark"], value="Light", label="Theme")
             feedback_radio = gr.Radio(
-                choices=["👍 Helpful", "👎 Not Helpful", "🔄 Needs Improvement"], 
+                choices=["👍 Helpful", "👎 Not Helpful", "🔄 Needs Improvement"],
                 label="Rate Last Response"
             )
         
         with gr.Row():
             with gr.Column():
                 follow_up_questions = gr.Radio(
-                    choices=[], 
+                    choices=[],
                     label="Suggested Follow-up Questions",
                     visible=False
                 )