#!/usr/bin/env python3
"""
Investigation report generator for Feed Hunter.
Reads triage.json and produces investigation tasks as structured prompts
for the agent to execute.

Usage:
    python3 investigate.py <triage.json> [--output investigations/]

This doesn't do the investigation itself — it generates the task list
that the agent (Case) follows using browser/web tools.
"""

import argparse
import json
import os
from datetime import datetime, timezone
from pathlib import Path


def generate_investigation_prompt(post):
    """Generate an investigation prompt for the agent."""
    author = post["author"].get("handle", "unknown")
    text = post["text"][:500]
    claims = post.get("claims", [])
    links = post.get("links", [])
    tasks = post.get("tasks", [])
    
    prompt = f"""## Investigation: {author}

**Post:** {text}

**Claims detected:**
"""
    for c in claims:
        prompt += f"- [{c['type']}] {c['match']}\n"
    
    prompt += f"\n**Links found:**\n"
    for l in links:
        prompt += f"- [{l['type']}] {l['url']}\n"
    
    prompt += f"\n**Investigation tasks:**\n"
    for i, t in enumerate(tasks, 1):
        prompt += f"{i}. **{t['action']}**: {t['description']}\n"
        prompt += f"   Method: {t['method']}\n"
        if t.get('url'):
            prompt += f"   URL: {t['url']}\n"
    
    prompt += """
**Deliver:**
1. Is the claim verifiable? What does the actual data show?
2. Is there recent activity? (Last 24-48h)
3. Is this still actionable or has the window closed?
4. Risk assessment (1-10, where 10 is highest risk)
5. Verdict: ACTIONABLE / EXPIRED / EXAGGERATED / SCAM / UNVERIFIABLE
6. If ACTIONABLE: suggested paper trade parameters (asset, entry, size, stop loss, take profit)
"""
    return prompt


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("input", help="Path to triage.json")
    parser.add_argument("--output", help="Output directory for investigation files")
    args = parser.parse_args()
    
    with open(args.input) as f:
        data = json.load(f)
    
    queue = data.get("investigation_queue", [])
    
    if not queue:
        print("No posts in investigation queue.")
        return
    
    output_dir = args.output or os.path.join(os.path.dirname(args.input), "investigations")
    os.makedirs(output_dir, exist_ok=True)
    
    timestamp = datetime.now(timezone.utc).strftime("%Y%m%d-%H%M%S")
    
    investigations = []
    for i, post in enumerate(queue):
        inv = {
            "id": f"inv-{timestamp}-{i}",
            "post_author": post["author"].get("handle", "unknown"),
            "post_url": post.get("url", ""),
            "priority": post["priority"],
            "claims": post.get("claims", []),
            "tasks": post.get("tasks", []),
            "prompt": generate_investigation_prompt(post),
            "status": "pending",
            "result": None,
        }
        investigations.append(inv)
    
    # Save investigation batch
    batch_file = os.path.join(output_dir, f"batch-{timestamp}.json")
    with open(batch_file, "w") as f:
        json.dump({
            "batch_id": timestamp,
            "created_at": datetime.now(timezone.utc).isoformat(),
            "count": len(investigations),
            "investigations": investigations,
        }, f, indent=2)
    
    print(f"=== Investigation Batch: {timestamp} ===")
    print(f"Tasks: {len(investigations)}")
    for inv in investigations:
        print(f"\n  [{inv['priority']}] {inv['post_author']}")
        print(f"  Claims: {[c['type'] for c in inv['claims']]}")
        print(f"  Tasks: {len(inv['tasks'])}")
    
    print(f"\nSaved to {batch_file}")
    print(f"\nTo execute: agent reads batch file and runs each investigation prompt")


if __name__ == "__main__":
    main()