app.py

"""
🌐 OpenCLAW Literary Agent — Autonomous 24/7 Orchestrator
==========================================================
Manages all platforms, generates content, engages with communities,
reflects on strategy, and self-improves continuously.

Runs on HuggingFace Spaces (Gradio dashboard) + GitHub Actions.
"""
import json
import logging
import os
import random
import threading
import time
from datetime import datetime, timezone, timedelta

# Setup logging
logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s [%(name)s] %(levelname)s: %(message)s",
    handlers=[logging.StreamHandler()]
)
logger = logging.getLogger("agent")

# Ensure state directory
STATE_DIR = os.environ.get("STATE_DIR", "state")
os.makedirs(STATE_DIR, exist_ok=True)
os.environ["STATE_DIR"] = STATE_DIR

from core import config
from core.content_engine import (
    research_post, novel_post, collaboration_post,
    fiction_research_bridge, agent_networking_post,
    engagement_reply, get_random_content
)
from platforms import moltbook, telegram_bot, reddit_bot, email_bot
from strategy.reflector import record_post, reflect, get_stats, get_current_strategy
from research.arxiv_fetcher import get_papers

# ── Agent State ──────────────────────────────────────────────────────

STATE_FILE = os.path.join(STATE_DIR, "agent_state.json")

def load_state() -> dict:
    try:
        if os.path.exists(STATE_FILE):
            with open(STATE_FILE) as f:
                return json.load(f)
    except Exception:
        pass
    return {
        "total_cycles": 0,
        "total_posts": 0,
        "total_engagements": 0,
        "last_post_time": None,
        "last_engagement_time": None,
        "last_reflection_time": None,
        "boot_time": datetime.now(timezone.utc).isoformat(),
        "log": [],
    }

def save_state(state: dict):
    os.makedirs(STATE_DIR, exist_ok=True)
    with open(STATE_FILE, "w") as f:
        json.dump(state, f, indent=2, default=str)

def log_action(state: dict, action: str, detail: str = ""):
    entry = {
        "time": datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M"),
        "action": action,
        "detail": detail[:200],
    }
    state["log"].append(entry)
    state["log"] = state["log"][-100:]
    logger.info(f"[{action}] {detail[:120]}")


# ── Platform Dispatcher ──────────────────────────────────────────────

def publish_to_platform(platform: str, content_data: dict) -> bool:
    """Publish content to a specific platform."""
    content = content_data.get("content", "")
    if not content:
        return False

    try:
        if platform == "moltbook" and moltbook.is_available():
            result = moltbook.post(content)
            return "error" not in result

        elif platform == "telegram" and telegram_bot.is_available():
            result = telegram_bot.post(content)
            return result.get("ok", False)

        elif platform == "reddit" and reddit_bot.is_available():
            title = content_data.get("paper", content_data.get("novel", "OpenCLAW Research Update"))
            if len(title) > 300:
                title = title[:297] + "..."
            subreddits = ["artificial", "MachineLearning", "OpenAI", "singularity"]
            sub = random.choice(subreddits)
            result = reddit_bot.post(title, content, sub)
            return "error" not in result

        else:
            # Log content for platforms without API (twitter, fb, linkedin, etc)
            logger.info(f"[{platform}] Content generated (no API): {content[:80]}...")
            return True

    except Exception as e:
        logger.error(f"Publish to {platform} failed: {e}")
        return False


# ── Autonomous Cycle ─────────────────────────────────────────────────

def run_post_cycle(state: dict) -> str:
    """Generate and publish content to a random platform."""
    platforms = ["moltbook", "telegram", "reddit", "twitter", "linkedin",
                 "facebook", "chirper", "instagram", "agentarxiv"]
    
    # Weight platforms by availability
    available = []
    for p in platforms:
        if p == "moltbook" and moltbook.is_available():
            available.extend([p] * 3)  # Higher weight
        elif p == "telegram" and telegram_bot.is_available():
            available.extend([p] * 2)
        elif p == "reddit" and reddit_bot.is_available():
            available.extend([p] * 2)
        else:
            available.append(p)

    platform = random.choice(available) if available else random.choice(platforms)
    
    try:
        content_data = get_random_content(platform)
        success = publish_to_platform(platform, content_data)
        
        record_post({
            "platform": platform,
            "type": content_data.get("type", "unknown"),
            "content_preview": content_data.get("content", "")[:100],
            "success": success,
        })
        
        state["total_posts"] += 1
        state["last_post_time"] = datetime.now(timezone.utc).isoformat()
        
        detail = f"{platform} | {content_data.get('type', '?')} | {'✅' if success else '❌'}"
        log_action(state, "POST", detail)
        return detail
    except Exception as e:
        log_action(state, "POST_ERROR", str(e))
        return f"Error: {e}"


def run_engagement_cycle(state: dict) -> str:
    """Engage with other agents/posts on Moltbook."""
    if not moltbook.is_available():
        return "Moltbook not available"
    
    try:
        results = moltbook.engage_with_hot_posts(engagement_reply)
        count = len(results)
        state["total_engagements"] += count
        state["last_engagement_time"] = datetime.now(timezone.utc).isoformat()
        detail = f"Engaged with {count} posts"
        log_action(state, "ENGAGE", detail)
        return detail
    except Exception as e:
        log_action(state, "ENGAGE_ERROR", str(e))
        return f"Error: {e}"


def run_reflection_cycle(state: dict) -> str:
    """Perform strategic self-reflection."""
    try:
        reflection = reflect()
        state["last_reflection_time"] = datetime.now(timezone.utc).isoformat()
        analysis = reflection.get("analysis", "")[:150]
        log_action(state, "REFLECT", analysis)
        return analysis
    except Exception as e:
        log_action(state, "REFLECT_ERROR", str(e))
        return f"Error: {e}"


# ── Main Autonomous Loop ─────────────────────────────────────────────

def autonomous_loop():
    """Run forever: post → engage → reflect → repeat."""
    state = load_state()
    log_action(state, "BOOT", f"Agent starting. Previous cycles: {state['total_cycles']}")
    
    # Send boot notification
    try:
        email_bot.send_boot_notification()
    except Exception:
        pass
    
    # Pre-fetch papers
    try:
        papers = get_papers()
        log_action(state, "ARXIV", f"Loaded {len(papers)} papers")
    except Exception:
        pass
    
    save_state(state)
    
    cycle = 0
    while True:
        try:
            cycle += 1
            state["total_cycles"] += 1
            now = datetime.now(timezone.utc)
            
            # ── POST (every cycle) ──
            post_result = run_post_cycle(state)
            logger.info(f"Cycle {cycle}: POST → {post_result}")
            
            # ── ENGAGE (every 2nd cycle) ──
            if cycle % 2 == 0:
                eng_result = run_engagement_cycle(state)
                logger.info(f"Cycle {cycle}: ENGAGE → {eng_result}")
            
            # ── REFLECT (every 6th cycle) ──
            if cycle % 6 == 0:
                ref_result = run_reflection_cycle(state)
                logger.info(f"Cycle {cycle}: REFLECT → {ref_result}")
            
            save_state(state)
            
            # Sleep between 3-6 hours (randomized)
            sleep_hours = random.uniform(
                config.POST_INTERVAL_MIN_HOURS,
                config.POST_INTERVAL_MAX_HOURS
            )
            sleep_secs = int(sleep_hours * 3600)
            logger.info(f"💤 Sleeping {sleep_hours:.1f}h until next cycle...")
            time.sleep(sleep_secs)
            
        except Exception as e:
            logger.error(f"Cycle error: {e}")
            log_action(state, "ERROR", str(e))
            save_state(state)
            time.sleep(300)  # 5 min on error


# ── CLI Mode (for GitHub Actions) ────────────────────────────────────

def run_once():
    """Run a single cycle (for GitHub Actions / cron)."""
    state = load_state()
    
    # Pre-fetch papers
    try:
        papers = get_papers()
        log_action(state, "ARXIV", f"Loaded {len(papers)} papers")
    except Exception:
        pass
    
    state["total_cycles"] += 1
    
    # Post to 2-3 platforms
    for _ in range(random.randint(2, 3)):
        run_post_cycle(state)
        time.sleep(5)
    
    # Engage
    run_engagement_cycle(state)
    
    # Reflect every 6th cycle
    if state["total_cycles"] % 6 == 0:
        run_reflection_cycle(state)
    
    save_state(state)
    
    # Print summary
    stats = get_stats()
    print(f"\n{'='*60}")
    print(f"🌐 OpenCLAW Literary Agent — Cycle #{state['total_cycles']}")
    print(f"   Total posts: {state['total_posts']}")
    print(f"   Total engagements: {state['total_engagements']}")
    print(f"   Posts by platform: {json.dumps(stats.get('by_platform', {}))}")
    print(f"   Posts by type: {json.dumps(stats.get('by_type', {}))}")
    print(f"{'='*60}\n")


# ── Gradio Dashboard (for HF Spaces) ─────────────────────────────────

def build_dashboard():
    """Build Gradio dashboard for monitoring and control."""
    try:
        import gradio as gr
    except ImportError:
        logger.warning("Gradio not installed, running CLI mode")
        return None

    def get_status():
        state = load_state()
        stats = get_stats()
        strategy = get_current_strategy()
        papers = get_papers()
        
        lines = [
            "# 🌐 OpenCLAW Literary Agent — Dashboard",
            "",
            f"**Status:** 🟢 ONLINE",
            f"**Boot time:** {state.get('boot_time', 'unknown')}",
            f"**Total cycles:** {state.get('total_cycles', 0)}",
            f"**Total posts:** {state.get('total_posts', 0)}",
            f"**Total engagements:** {state.get('total_engagements', 0)}",
            "",
            "## 📊 Posts by Platform",
        ]
        for plat, count in stats.get("by_platform", {}).items():
            lines.append(f"- **{plat}**: {count}")
        
        lines.extend([
            "", "## 📝 Posts by Type",
        ])
        for typ, count in stats.get("by_type", {}).items():
            lines.append(f"- **{typ}**: {count}")
        
        lines.extend([
            "", f"## 📄 ArXiv Papers Loaded: {len(papers)}",
        ])
        for p in papers[:5]:
            lines.append(f"- [{p.get('id','')}] {p.get('title','')[:60]}")
        
        lines.extend([
            "", "## 📡 Platform Status",
            f"- Moltbook: {'🟢' if moltbook.is_available() else '🔴'}",
            f"- Telegram: {'🟢' if telegram_bot.is_available() else '🔴'}",
            f"- Reddit: {'🟢' if reddit_bot.is_available() else '🔴'}",
            f"- Email: {'🟢' if email_bot.is_available() else '🔴'}",
            f"- LLM (NVIDIA): {'🟢' if config.NVIDIA_KEYS else '🔴'}",
            f"- LLM (Groq): {'🟢' if config.GROQ_KEYS else '🔴'}",
            f"- LLM (Gemini): {'🟢' if config.GEMINI_KEYS else '🔴'}",
        ])
        
        return "\n".join(lines)

    def get_recent_log():
        state = load_state()
        log = state.get("log", [])
        if not log:
            return "No activity yet."
        lines = ["| Time | Action | Detail |", "|------|--------|--------|"]
        for entry in reversed(log[-25:]):
            lines.append(f"| {entry.get('time','')} | {entry.get('action','')} | {entry.get('detail','')[:80]} |")
        return "\n".join(lines)

    def trigger_post():
        state = load_state()
        result = run_post_cycle(state)
        save_state(state)
        return f"✅ {result}\n\n{get_recent_log()}"

    def trigger_engage():
        state = load_state()
        result = run_engagement_cycle(state)
        save_state(state)
        return f"✅ {result}\n\n{get_recent_log()}"

    def trigger_reflect():
        state = load_state()
        result = run_reflection_cycle(state)
        save_state(state)
        return f"✅ {result}"

    with gr.Blocks(title="OpenCLAW Literary Agent", theme=gr.themes.Soft()) as app:
        gr.Markdown("# 🌐 OpenCLAW Literary Agent — Autonomous 24/7")
        gr.Markdown(f"**Author:** {config.AUTHOR_NAME} | [Wikipedia]({config.AUTHOR_WIKIPEDIA}) | [GitHub]({config.AUTHOR_GITHUB}) | [Scholar]({config.AUTHOR_SCHOLAR})")
        
        with gr.Row():
            with gr.Column(scale=2):
                status_md = gr.Markdown(get_status())
                refresh_btn = gr.Button("🔄 Refresh Status", variant="secondary")
                refresh_btn.click(fn=get_status, outputs=status_md)
            
            with gr.Column(scale=1):
                gr.Markdown("### ⚡ Manual Controls")
                post_btn = gr.Button("📝 Publish Now", variant="primary")
                engage_btn = gr.Button("💬 Engage Now", variant="primary")
                reflect_btn = gr.Button("🧠 Reflect Now", variant="secondary")
                manual_output = gr.Markdown("")
                
                post_btn.click(fn=trigger_post, outputs=manual_output)
                engage_btn.click(fn=trigger_engage, outputs=manual_output)
                reflect_btn.click(fn=trigger_reflect, outputs=manual_output)
        
        with gr.Row():
            log_md = gr.Markdown(get_recent_log())
            log_btn = gr.Button("🔄 Refresh Log")
            log_btn.click(fn=get_recent_log, outputs=log_md)

    return app


# ── Entry Point ──────────────────────────────────────────────────────

if __name__ == "__main__":
    import sys
    
    if len(sys.argv) > 1 and sys.argv[1] == "once":
        # CLI mode for GitHub Actions
        run_once()
    else:
        # Dashboard + background agent
        app = build_dashboard()
        if app:
            # Start autonomous loop in background
            bg = threading.Thread(target=autonomous_loop, daemon=True)
            bg.start()
            # Launch Gradio
            app.launch(server_name="0.0.0.0", server_port=7860, share=False)
        else:
            # No Gradio, run autonomous loop directly
            autonomous_loop()