oai.py

@@ -4532,12 +4532,24 @@ def chat():
                 current_index = len(session_history) - 1
                 if conversation_memory_enabled:
                     memory_start_index = 0
+                
+                # Recalculate session totals from loaded message history
                 total_input_tokens = 0
                 total_output_tokens = 0
                 total_cost = 0.0
                 message_count = 0
+                
+                for msg in session_history:
+                    # Handle both old format (no cost data) and new format (with cost data)
+                    total_input_tokens += msg.get('prompt_tokens', 0)
+                    total_output_tokens += msg.get('completion_tokens', 0)
+                    total_cost += msg.get('msg_cost', 0.0)
+                    message_count += 1
+                
                 console.print(f"[bold green]Conversation '{conversation_name}' loaded with {len(session_history)} messages.[/]")
-                app_logger.info(f"Conversation '{conversation_name}' loaded with {len(session_history)} messages")
+                if total_cost > 0:
+                    console.print(f"[dim cyan]Restored session totals: {total_input_tokens + total_output_tokens} tokens, ${total_cost:.4f} cost[/]")
+                app_logger.info(f"Conversation '{conversation_name}' loaded with {len(session_history)} messages, restored cost: ${total_cost:.4f}")
                 continue
             
             elif user_input.lower().startswith("/delete"):
@@ -4680,8 +4692,8 @@ def chat():
                 total_output_tokens = 0
                 total_cost = 0.0
                 message_count = 0
-                console.print("[bold green]Conversation context reset.[/]")
+                console.print("[bold green]Conversation context reset. Totals cleared.[/]")
-                app_logger.info("Conversation context reset by user")
+                app_logger.info("Conversation context reset by user - all totals reset to 0")
                 continue
 
             elif user_input.lower().startswith("/info"):
@@ -5776,10 +5788,13 @@ All queries are read-only. INSERT/UPDATE/DELETE are not allowed."""
             stream_interrupted = False
 
             if is_streaming:
+                # Store the last chunk to get usage data after streaming completes
+                last_chunk = None
                 try:
                     with Live("", console=console, refresh_per_second=10, auto_refresh=True) as live:
                         try:
                             for chunk in response:
+                                last_chunk = chunk  # Keep track of last chunk for usage data
                                 if hasattr(chunk, 'error') and chunk.error:
                                     console.print(f"\n[bold red]Stream error: {chunk.error.message}[/]")
                                     app_logger.error(f"Stream error: {chunk.error.message}")
@@ -5835,6 +5850,14 @@ All queries are read-only. INSERT/UPDATE/DELETE are not allowed."""
                         pass
     
                     continue  # Now it's safe to continue
+                
+                # For streaming, try to get usage from last chunk
+                if last_chunk and hasattr(last_chunk, 'usage'):
+                    response.usage = last_chunk.usage
+                    app_logger.debug("Extracted usage data from last streaming chunk")
+                elif last_chunk:
+                    app_logger.warning("Last streaming chunk has no usage data")
+                
             else:
                 full_response = response.choices[0].message.content if response.choices else ""
                 # Clear any processing messages before showing response
@@ -5847,13 +5870,54 @@ All queries are read-only. INSERT/UPDATE/DELETE are not allowed."""
                     console.print()
                     console.print(Panel(md, title="[bold green]AI Response[/]", title_align="left", border_style="green"))
                 
-                session_history.append({'prompt': user_input, 'response': full_response})
+                # Extract usage data BEFORE appending to history
-                current_index = len(session_history) - 1
-
                 usage = getattr(response, 'usage', None)
-                input_tokens = usage.input_tokens if usage and hasattr(usage, 'input_tokens') else 0
+                
-                output_tokens = usage.output_tokens if usage and hasattr(usage, 'output_tokens') else 0
+                # DEBUG: Log what OpenRouter actually returns
-                msg_cost = usage.total_cost_usd if usage and hasattr(usage, 'total_cost_usd') else estimate_cost(input_tokens, output_tokens)
+                if usage:
+                    app_logger.debug(f"Usage object type: {type(usage)}")
+                    app_logger.debug(f"Usage attributes: {dir(usage)}")
+                    if hasattr(usage, '__dict__'):
+                        app_logger.debug(f"Usage dict: {usage.__dict__}")
+                else:
+                    app_logger.warning("No usage object in response!")
+                
+                # Try both attribute naming conventions (OpenAI standard vs Anthropic)
+                input_tokens = 0
+                output_tokens = 0
+                
+                if usage:
+                    # Try prompt_tokens/completion_tokens (OpenAI/OpenRouter standard)
+                    if hasattr(usage, 'prompt_tokens'):
+                        input_tokens = usage.prompt_tokens or 0
+                    elif hasattr(usage, 'input_tokens'):
+                        input_tokens = usage.input_tokens or 0
+                    
+                    if hasattr(usage, 'completion_tokens'):
+                        output_tokens = usage.completion_tokens or 0
+                    elif hasattr(usage, 'output_tokens'):
+                        output_tokens = usage.output_tokens or 0
+                
+                app_logger.debug(f"Extracted tokens: input={input_tokens}, output={output_tokens}")
+                
+                # Get cost from API or estimate
+                msg_cost = 0.0
+                if usage and hasattr(usage, 'total_cost_usd') and usage.total_cost_usd:
+                    msg_cost = float(usage.total_cost_usd)
+                    app_logger.debug(f"Using API cost: ${msg_cost:.6f}")
+                else:
+                    msg_cost = estimate_cost(input_tokens, output_tokens)
+                    app_logger.debug(f"Estimated cost: ${msg_cost:.6f} (from {input_tokens} input + {output_tokens} output tokens)")
+                
+                # NOW append to history with cost data
+                session_history.append({
+                    'prompt': user_input,
+                    'response': full_response,
+                    'msg_cost': msg_cost,
+                    'prompt_tokens': input_tokens,
+                    'completion_tokens': output_tokens
+                })
+                current_index = len(session_history) - 1
 
                 total_input_tokens += input_tokens
                 total_output_tokens += output_tokens