Keep GitHub forks in sync at scale. Supports 10,000+ repos. Runs in seconds.

1. Every night at 6am UTC, forksync checks which forks are due based on upstream activity
2. Forks are synced using gh repo sync — the official GitHub CLI tool
3. Every sync is verified — a fork is only marked synced if confirmed up-to-date
4. A SYNC_REPORT.md is committed showing exactly what happened

forksync checks repos based on how active their upstream is:

At 10,000 repos this reduces nightly API calls by 70-80% compared to checking everything every night.

GitHub Actions (nightly cron)
  → POST /run to forksync Cloud Run service

forksync service
  → GraphQL batch fetch all fork metadata (9 calls for 800 forks)
  → Firestore batch read existing fork state
  → Filter to only forks due for checking (schedule tiers)
  → Compare API concurrently (50 parallel) with ETag caching
  → gh repo sync concurrently (20 parallel) for behind forks
  → Verify each sync with compare API
  → Firestore batch write results
  → Commit SYNC_REPORT.md

• GCP project with Firestore enabled
• Cloud Run service deployed
• GitHub token with repo scope

gcloud run deploy forksync \
  --source . \
  --region us-central1 \
  --set-env-vars GH_TOKEN=your_token,FORK_OWNER=your_username

Go to Settings → Secrets → Actions:

• FORKSYNC_SERVICE_URL — your Cloud Run service URL
• FORKSYNC_API_KEY — API key for the service

GITHUB_TOKEN is provided automatically by GitHub Actions.

Go to Actions tab → Enable workflows

Syncing starts tonight at 6am UTC.

Create fork-sync.yml:

sync:
  fork_owner: your-github-username
  concurrency_compare: 50
  concurrency_sync: 20
  verify_after_sync: true

schedule:
  nightly_threshold_days: 30
  weekly_threshold_days: 365

gcp:
  project_id: your-project-id
  firestore_collection: forks

notifications:
  slack_webhook: ""
  discord_webhook: ""

pip install forksync

# Dry run
python -m forksync run --dry-run

# Live sync
python -m forksync run

# Check status
python -m forksync status

# Sync specific repos
python -m forksync run --repos vllm langchain

# View sync history
python -m forksync history

• Fast-forward only — never syncs a fork where you're ahead of upstream
• No force push — ever
• Verified syncs — every sync confirmed via compare API post-sync
• Conflict issues — diverged forks get a GitHub Issue
• Dry run support — preview without making changes

forksync is pip-installable and can be used as a library:

from forksync import run_sync, get_status

# Run a full sync
results = await run_sync()

# Get status of all forks
forks = await get_status()

Reporium uses forksync as a library to keep its fork registry current.

Perditio · Part of the Reporium suite