#!/usr/bin/env python3

import json
import os
import sys
import glob
from collections import defaultdict
from datetime import datetime

def extract_text_content(content_list):
    """Extract text content from message content array, excluding thinking blocks."""
    text_parts = []
    for item in content_list:
        if item.get('type') == 'text':
            text_parts.append(item['text'])
    return ' '.join(text_parts).strip()

def process_session_file(file_path):
    """Process a single JSONL session file and extract conversation turns."""
    turns = []
    
    try:
        with open(file_path, 'r') as f:
            lines = f.readlines()
        
        messages = []
        for line in lines:
            try:
                data = json.loads(line.strip())
                if data.get('type') == 'message' and 'message' in data:
                    messages.append(data)
            except json.JSONDecodeError:
                continue
        
        # Group messages into conversation turns (user -> assistant pairs)
        current_user_msg = None
        
        for msg in messages:
            role = msg['message']['role']
            content = msg['message'].get('content', [])
            
            if role == 'user':
                current_user_msg = extract_text_content(content)
            elif role == 'assistant' and current_user_msg:
                assistant_text = extract_text_content(content)
                if assistant_text and current_user_msg:  # Only include if both have content
                    turns.append({
                        'user': current_user_msg,
                        'assistant': assistant_text,
                        'timestamp': msg.get('timestamp')
                    })
                current_user_msg = None
        
        return turns
    
    except Exception as e:
        print(f"Error processing {file_path}: {e}", file=sys.stderr)
        return []

def main():
    sessions_dir = "/home/wdjones/.openclaw/agents/main/sessions/"
    
    # Get all non-deleted session files
    session_files = []
    for file_path in glob.glob(os.path.join(sessions_dir, "*.jsonl")):
        if not file_path.endswith('.deleted'):
            session_files.append(file_path)
    
    # Sort by modification time (newest first)
    session_files.sort(key=os.path.getmtime, reverse=True)
    
    all_turns = []
    
    # Process session files until we have enough turns
    for session_file in session_files:
        turns = process_session_file(session_file)
        all_turns.extend(turns)
        
        if len(all_turns) >= 10:
            break
    
    # Get the last 10 turns
    recent_turns = all_turns[-10:] if len(all_turns) > 10 else all_turns
    
    # Output the conversation turns as JSON for the auto-memory script
    for turn in recent_turns:
        output = {
            "user": turn['user'],
            "assistant": turn['assistant'],
            "agent_id": "case",
            "session": "main"
        }
        print(json.dumps(output))

if __name__ == "__main__":
    main()