agent-skills/run.py

#!/usr/bin/env python3
"""Agent Skills Pipeline Runner — full 8-stage pipeline."""
import yaml
import json
import sys
import os
import datetime

sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))

from pipeline.scout import scout
from pipeline.filter import filter_repos
from pipeline.reader import read_repo
from pipeline.extractor import extract_workflow
from pipeline.scorer import score_workflow
from pipeline.generator import generate_skill
from pipeline.reviewer import review_skill
from pipeline.publisher import publish_skill

def load_config():
    config_path = os.path.join(os.path.dirname(__file__), "config", "settings.yaml")
    with open(config_path) as f:
        return yaml.safe_load(f)

def main():
    config = load_config()
    runs_dir = os.path.join(os.path.dirname(__file__), "runs")
    os.makedirs(runs_dir, exist_ok=True)

    run_id = datetime.datetime.now().strftime("%Y%m%d-%H%M%S")
    run_log = {"run_id": run_id, "started_at": datetime.datetime.now().isoformat(), "stages": {}}

    print("=" * 60)
    print("Agent Skills Pipeline — Run", run_id)
    print("=" * 60)

    # --- Stage 1: Scout ---
    print("\n[1/8] Scout — Discovering repos from GitHub...")
    state = {}
    scout_result = scout(config, state)

    if scout_result.get("status") in ("COOLDOWN", "RATE_LIMITED"):
        print(f"  ⏸ {scout_result.get('message', scout_result.get('status'))}")
        return

    repos = scout_result.get("repos", [])
    print(f"  Found {scout_result.get('count', 0)} repos")
    run_log["stages"]["scout"] = {"count": scout_result.get("count", 0)}

    if not repos:
        print("  No repos found. Exiting.")
        return

    # --- Stage 2: Filter ---
    print("\n[2/8] Filter — Applying deterministic rules...")
    filter_result = filter_repos(repos, config)
    kept = filter_result.get("kept", [])
    print(f"  Kept: {filter_result.get('kept_count', 0)} | Rejected: {filter_result.get('rejected_count', 0)}")
    run_log["stages"]["filter"] = {"kept": filter_result.get("kept_count", 0), "rejected": filter_result.get("rejected_count", 0)}

    for rej in filter_result.get("rejected", [])[:3]:
        print(f"    ✗ {rej['repo'].get('name', '?')}: {', '.join(rej['reasons'])}")

    if not kept:
        print("  All repos filtered out. Exiting.")
        return

    # --- Stages 3-8: Process each kept repo ---
    print(f"\n[3-8/8] Processing {len(kept)} repos through pipeline...")
    results = {"extracted": 0, "scored": 0, "generated": 0, "reviewed": 0, "published": 0}

    for i, repo in enumerate(kept[:5]):  # Cap at 5 per run
        repo_name = repo.get("full_name", repo.get("name", "?"))
        repo_url = repo.get("url", "")
        print(f"\n  ── Repo {i+1}/{min(len(kept), 5)}: {repo_name} ──")

        # Stage 3: Reader
        print(f"    [3/8] Reader — Loading context...")
        reader_output = read_repo(repo.get("clone_url", repo_url), config)

        if reader_output.get("status") == "INSUFFICIENT":
            print(f"    ⏸ Insufficient context: {reader_output.get('decision_reason', '')}")
            continue

        print(f"    Loaded {len(reader_output.get('context_loaded', []))} files")

        # Stage 4: Extractor
        print(f"    [4/8] Extractor — Looking for reusable workflow...")
        extract_output = extract_workflow(reader_output, config)

        if extract_output.get("status") != "EXTRACTED":
            print(f"    ✗ No workflow: {extract_output.get('reason', extract_output.get('status', ''))}")
            continue

        workflow = extract_output.get("workflow", {})
        print(f"    ✓ Extracted: {workflow.get('skill_name', '?')} (confidence: {workflow.get('confidence', 0)})")
        results["extracted"] += 1

        # Stage 5: Scorer
        print(f"    [5/8] Scorer — Evaluating...")
        extract_output["reader_output"] = reader_output
        score_output = score_workflow(extract_output, config)

        if score_output.get("decision") != "PASS":
            print(f"    ✗ Score {score_output.get('score', 0)} < {score_output.get('min_score', 0.85)}")
            continue

        print(f"    ✓ Score: {score_output.get('score', 0)} (passed)")
        results["scored"] += 1

        # Stage 6: Generator
        print(f"    [6/8] Generator — Building Skill package...")
        gen_output = generate_skill(score_output, config)

        if gen_output.get("status") != "GENERATED":
            print(f"    ✗ Generation blocked: {gen_output.get('reason', '')}")
            continue

        print(f"    ✓ Generated: {gen_output.get('skill_name', '?')} ({len(gen_output.get('files', {}))} files)")
        results["generated"] += 1

        # Stage 7: Reviewer
        print(f"    [7/8] Reviewer — LLM review...")
        review_output = review_skill(gen_output, config)

        if review_output.get("status") != "APPROVED":
            print(f"    ✗ Review: {review_output.get('status', '?')} — {review_output.get('reason', '')[:100]}")
            continue

        print(f"    ✓ Approved (confidence: {review_output.get('confidence', 0)})")
        results["reviewed"] += 1

        # Stage 8: Publisher
        print(f"    [8/8] Publisher — Creating PR...")
        publish_output = publish_skill(review_output, config)

        if publish_output.get("status") == "PUBLISHED":
            print(f"    ✓ Published! PR: {publish_output.get('pr_url', '')}")
            results["published"] += 1
        else:
            print(f"    ! {publish_output.get('status', '?')}: {publish_output.get('message', publish_output.get('error', ''))[:100]}")

    # --- Summary ---
    print("\n" + "=" * 60)
    print("PIPELINE COMPLETE")
    print("=" * 60)
    print(f"  Scout:     {scout_result.get('count', 0)} discovered")
    print(f"  Filter:    {filter_result.get('kept_count', 0)} kept / {filter_result.get('rejected_count', 0)} rejected")
    print(f"  Extracted: {results['extracted']}")
    print(f"  Scored:    {results['scored']}")
    print(f"  Generated: {results['generated']}")
    print(f"  Reviewed:  {results['reviewed']}")
    print(f"  Published: {results['published']}")

    # Save run log
    run_log["results"] = results
    run_log["ended_at"] = datetime.datetime.now().isoformat()
    log_path = os.path.join(runs_dir, f"{run_id}.json")
    with open(log_path, 'w') as f:
        json.dump(run_log, f, indent=2)
    print(f"\n  Run log: {log_path}")

if __name__ == "__main__":
    main()