Skip to content

Latest commit

 

History

History

app

README.md

openlaunch.lol — app

Launch a token. Free. Open source. On Base, Robinhood Chain or Arc.

The site for the openlaunch launchpad. One transaction deploys a fixed-supply ERC-20, opens a Uniswap v4 pool and locks 100% of the supply as liquidity in an ownerless locker — forever. No platform fee anywhere. Contracts live in ../contracts (LaunchFactory.sol, LaunchLocker.sol, LaunchToken.sol; design in ../contracts/docs/LAUNCHPAD.md).

Next 16 app router, React 19, Tailwind 4, wagmi 3 + viem 2, Postgres (postgres.js). Runs on Fly.io (two web machines + an attached Postgres with WAL backups to Tigris); see LAUNCH.md for the runbook.

Pages / API

  • / list (new · active · volume · market cap) + live tape · /launch form · /t/<token> trade + fees
  • /rules how it works · /agents contract call + JSON · /llms.txt
  • GET /api/launch/list?sort=&limit= · GET /api/launch/feed · GET /api/launch/meta/<token>
  • POST /api/launch/meta (metadata before launching → {uri, token}) · POST /api/launch/sync?tx= · GET /api/health

How it hangs together

  • src/lib/launchpad/ — abi.ts (factory / locker / PoolManager Swap / StateView / Quoter / Universal Router / Permit2), math.ts (ticks ↔ prices, FDV, formatting), swap.ts (V4_SWAP encoding), meta.ts, queries.ts, indexer.ts (receipt apply + chunked poller), loop.ts.
  • Trades happen in the page through the Universal Router (TradePanel.tsx): buys send ETH as value; sells approve the token to Permit2 once, then a 30-day Permit2 allowance to the router.
  • db/schema.sql — bb_launches, bb_launch_swaps, bb_launch_fee_events, bb_launch_meta, bb_launch_sync_state (+ bb_migrations). Idempotent; applied on every deploy.

Local dev against a Base fork

anvil --fork-url https://mainnet.base.org --port 8545 --chain-id 8453
cd ../contracts && DEPLOYER_PRIVATE_KEY=<anvil key 0> forge script script/dev/SeedLaunchpadLocal.s.sol --rpc-url http://127.0.0.1:8545 --broadcast
createdb basebid_dev && DATABASE_URL=postgres://localhost/basebid_dev npm run migrate
# .env.development.local: NEXT_PUBLIC_CHAIN=base, NEXT_PUBLIC_RPC_URL/BASE_RPC_URL=http://127.0.0.1:8545,
#   NEXT_PUBLIC_LAUNCH_FACTORY/_LOCKER + LAUNCH_DEPLOY_BLOCK from the seed output, LAUNCH_SYNC_CONFIRMATIONS=0, LAUNCH_SYNC_LOOP=1
npm run dev

Gotcha: anvil's default accounts carry EIP-7702 delegations on Base mainnet — ETH sent to them is forwarded away. Use a fresh key (cast wallet new + anvil_setBalance) when testing sells.

Deploy

First-time setup (secrets, storage, contracts) is in LAUNCH.md. Every deploy after that is a tagged release:

Releasing

Releases are cut by release-please from the conventional-commit titles of the pull requests merged into main (they are squash-merged, so each title is one commit). The version lives in package.json (SemVer: fix is a patch, feat a minor version, ! a major one); the notes go to CHANGELOG.md and GitHub Releases. The site footer and /api/health show the running version and commit.

  1. After each merge, release-please keeps one pull request titled chore: release X.Y.Z up to date: the version in package.json and package-lock.json, and the new CHANGELOG.md entry. Edit the notes there if needed; they are public: no hostnames, secrets, wallet keys or unfixed vulnerabilities.
  2. A maintainer closes and reopens the release pull request, right away. For what release-please's token does, GitHub creates no pr-guard run at all and holds the other checks for approval (seen on #150); a reopen by a maintainer is their own event, so every check runs, pr-guard included. Before merging, check that the pull request still carries its autorelease: pending label (release-please finds the merged release pull request by it; add it back if it is missing). Then the maintainer approves the pull request (pr-guard asks for an owner's approval), re-runs pr-guard (it reads approvals only when it runs, see CONTRIBUTING.md), and squash-merges it once it passes. release-please tags the merge commit vX.Y.Z and publishes the GitHub Release.
  3. Deploy, by hand, from an up-to-date main (git pull), at the repository root: app/scripts/deploy.sh --dry-run, then app/scripts/deploy.sh, both through the maintainers' checker below. deploy.sh refuses anything but a clean main equal to origin/main whose tag is on GitHub and whose release is published; it takes the contract addresses from the hash-pinned deployment records and checks that production reports the new version and commit. Then run app/scripts/smoke.sh.

A bad release is fixed by a new patch release, never by moving or deleting a tag. If release-please is ever off, a release can still be made by hand: in one pull request bump the version in package.json and package-lock.json and the app entry in .release-please-manifest.json to the same number (release-please counts from that manifest), tag the merge commit, publish the release with its notes, then deploy as above.

The release tooling (scripts/deploy.sh, smoke.sh, migrate.mjs and scripts/lib/, the Dockerfile, .dockerignore, fly.toml, and release-please-config.json at the repository root) is blocked by pr-guard for everyone: it runs with the deploy credentials or against the production database, or decides how releases are cut, so it changes only in a pull request of its own that a maintainer merges with the guard bypassed. Maintainers also deploy through a checker kept outside the repository, which compares these files with the fingerprints of the last reviewed release before it runs deploy.sh.