"""Pull issues + PRs + comments from a GitHub repo into JSONL. Usage: python scripts/pull_gh.py --repo huggingface/peft --months 6 --out data/raw Env: set GITHUB_TOKEN for 5000 req/hr (else 60 req/hr unauth). """ from __future__ import annotations import argparse import json import os import sys import time from datetime import datetime, timedelta, timezone from pathlib import Path import requests API = "https://api.github.com" def gh_get(url: str, token: str | None, params: dict | None = None) -> dict | list: headers = {"Accept": "application/vnd.github+json", "X-GitHub-Api-Version": "2022-11-28"} if token: headers["Authorization"] = f"Bearer {token}" for attempt in range(5): r = requests.get(url, headers=headers, params=params, timeout=30) if r.status_code == 200: return r.json() if r.status_code in (403, 429): reset = int(r.headers.get("X-RateLimit-Reset", "0")) wait = max(5, reset - int(time.time()) + 2) if reset else 30 print(f" rate-limited; sleeping {wait}s", flush=True) time.sleep(min(wait, 90)) continue if r.status_code == 404: return [] print(f" HTTP {r.status_code}: {r.text[:200]}", flush=True) time.sleep(2 + attempt) raise RuntimeError(f"failed after 5 attempts: {url}") def paginate(url: str, token: str | None, params: dict, page_cap: int = 50) -> list: out: list = [] page = 1 while page <= page_cap: p = dict(params, page=page, per_page=100) chunk = gh_get(url, token, p) if not chunk: break out.extend(chunk) if len(chunk) < 100: break page += 1 return out def pull_repo(repo: str, since_iso: str, token: str | None, out_dir: Path) -> dict: out_dir.mkdir(parents=True, exist_ok=True) print(f"[pull] {repo} since {since_iso}", flush=True) issues = paginate( f"{API}/repos/{repo}/issues", token, {"state": "all", "since": since_iso, "sort": "created", "direction": "desc"}, ) print(f" fetched {len(issues)} issues+prs", flush=True) (out_dir / "issues.jsonl").write_text( "\n".join(json.dumps(x) for x in issues), encoding="utf-8" ) labels = gh_get(f"{API}/repos/{repo}/labels", token, {"per_page": 100}) if isinstance(labels, list): (out_dir / "labels.jsonl").write_text( "\n".join(json.dumps(x) for x in labels), encoding="utf-8" ) print(f" fetched {len(labels)} labels", flush=True) comments_count = 0 with open(out_dir / "comments.jsonl", "w", encoding="utf-8") as cf: for i, it in enumerate(issues): if it.get("comments", 0) == 0: continue num = it["number"] cs = paginate( f"{API}/repos/{repo}/issues/{num}/comments", token, {}, page_cap=5, ) for c in cs: c["_issue_number"] = num cf.write(json.dumps(c) + "\n") comments_count += 1 if i % 25 == 0: print(f" {i}/{len(issues)} issues, {comments_count} comments so far", flush=True) print(f" fetched {comments_count} comments total", flush=True) return {"issues": len(issues), "comments": comments_count} def main(): ap = argparse.ArgumentParser() ap.add_argument("--repo", default="huggingface/peft") ap.add_argument("--months", type=int, default=6) ap.add_argument("--out", default="data/raw") args = ap.parse_args() token = os.environ.get("GITHUB_TOKEN") if not token: print("WARN: no GITHUB_TOKEN; rate limit = 60/hr. Set token for 5000/hr.", flush=True) since = (datetime.now(timezone.utc) - timedelta(days=30 * args.months)).strftime("%Y-%m-%dT%H:%M:%SZ") out_dir = Path(args.out) / args.repo.replace("/", "_") stats = pull_repo(args.repo, since, token, out_dir) print(f"DONE: {stats}", flush=True) if __name__ == "__main__": sys.exit(main())