fairwins-infra

name

fairwins-infra

description

Bring the FairWins gasless Cloud Run infrastructure (the alto ERC-4337 bundler and the relay/paymaster signer gateway on Polygon) up or down to control cost, and check its status, health, and config. Use whenever the user wants to start/stop/turn on/turn off the bundler, gateway, paymaster, relayer, or "signers"; ask whether they're running; scale them to zero to save money; or bring them up before testing passkey gasless transactions. The SPA runs itself and is out of scope.

FairWins gasless infra control

The FairWins app (SPA) runs cheaply on its own and scales to zero. Two always-on Cloud Run services back the gasless (passkey / ERC-4337) flows and are the expensive part — each runs min-instances=1 with cpu-throttling=false, i.e. a full vCPU allocated 24/7. When we aren't actively testing or serving gasless traffic, scale them to zero; bring them back up on demand.

Alias	Cloud Run service	Role
bundler	fairwins-alto-bundler	ERC-4337 bundler (alto) — submits UserOperations
gateway	fairwins-relay-gateway	ERC-7677 paymaster (sponsorship signer) + EIP-3009 relay

Both are in project chippr-bots-site-wp, region us-central1.

Out of scope — never touch these: the SPA (prediction-dao-research, already scales to zero) and other projects' services (clearpath-*, fukuii-*, kings-edge-*). The KMS signing keys are cheap and must never be deleted — this skill only toggles the Cloud Run services that use them.

How to run

The script is manage.sh next to this file. Run it from the repo root:

bash .claude/skills/fairwins-infra/manage.sh status          # both services: state + health + bundler config
bash .claude/skills/fairwins-infra/manage.sh up              # warm both (min=1) + health check
bash .claude/skills/fairwins-infra/manage.sh down            # scale both to zero (min=0)
bash .claude/skills/fairwins-infra/manage.sh up bundler      # just the bundler
bash .claude/skills/fairwins-infra/manage.sh down gateway    # just the gateway

• down sets min-instances=0. The service scales to zero and idle cost drops to ~$0; it still cold-starts (a few seconds) to serve an on-demand request. This is the cost-saving state.
• up sets min-instances=1 (kept warm) and prints a health check. Use before an active testing session or when serving real gasless traffic, so users don't eat cold starts.
• status shows each service's current scale state, Ready condition, a live health ping, and — for the bundler — its critical env config.

Scaling changes do not alter env vars, so up/down are safe and reversible.

Typical flow

1. ... up before a gasless test session → wait for healthy → tell the user to go.
2. ... down when done → confirms scale-to-zero.
3. ... status any time to answer "is the bundler running?" / "are the signers up?".

Config-drift guard (important for this project)

status/up also read the bundler's env and warn if it drifted, because an automated deploy (the compute service account applying the repo service.yaml) has silently reverted the working config mid-session. The correct live config is:

• ALTO_RPC_URL → a QuickNode endpoint (NOT publicnode — that 403s archive reads and breaks receipts)
• ALTO_DEPLOY_SIMULATIONS_CONTRACT → true (else bundle-build fails before broadcast; executor nonce freezes)
• ALTO_GAS_PRICE_MULTIPLIERS → set, e.g. 400,500,600 (public RPCs report a stale ~30 gwei priority; Polygon congestion needs 100+ gwei)

If the guard reports drift, the fix is not in this skill — redeploy those env vars (or merge PR #895, which makes the manifest match and stops the clobber). See the polygon-gasless-bundler-stall memory for the full history.

Notes

• Requires an authenticated gcloud with access to chippr-bots-site-wp.
• Health URLs: bundler https://bundler.fairwins.app (JSON-RPC eth_supportedEntryPoints), gateway https://relay.fairwins.app (any HTTP response = reachable; it's origin-locked so a 404 on / is normal).
• The executor gas wallet (0x7C6d…) and the paymaster deposit are funded separately; scaling to zero does not affect on-chain balances.