STAMPD
KOL call → market → onchain receipt. Crypto calls posted on X become YES/NO prediction markets on Robinhood Chain, traded with demo USD and settled onchain with evidence. Every KOL builds a track record: hit rate and edge (how much they beat the crowd), always shown with their sample size.
Plans: xkol-analysis-and-web3-plan.md (the "plan file") and development-plan.md. Decisions made while building: docs/decisions.md.
Layout
apps/
web/ Next.js 16 · React 19 · Tailwind 4 · wagmi/RainbowKit + SIWE — every page in dev plan 1.7
api/ Fastify 5 + Zod — markets, KOLs, portfolio, faucet vouchers, submissions, notifications (SSE), admin
worker/ BullMQ — ingest, AI drafting, market creation, indexer, resolver bot, stats, notifications
scraper/ Python 3.12 · FastAPI · twscrape (pinned) — the only X data source (plan B6b)
packages/
contracts/ Foundry · OpenZeppelin v5 — DemoUSD, OutcomeTokens, MarketHub (FPMM), MarketFactory, Resolver
chain/ Robinhood Chain definitions (4663 / 46630), ABIs, deployments, clients
core/ FPMM quote math (matches Solidity to the unit), edge score, formatting, question hash
ai/ DeepSeek extraction + check passes, prefilter, validators, moderation routing
db/ Prisma 7 schema + migration (27 tables), PGlite test database
queue/ Job names/payloads shared by api and worker
ui/ "Neon Receipt" design tokens (D14), components, adapted React Bits (MIT + Commons Clause — never publish)
docs/ decisions · runbook · security · moderation playbook
Quick start (no Docker, no Redis)
Needs Node 22+, pnpm 11, Foundry and, for the scraper, uv.
pnpm install
pnpm dev:local # PGlite + anvil + contracts + indexer + API, one sample market
NEXT_PUBLIC_CHAIN_ID=31337 pnpm --filter @stampd/web dev # http://localhost:3000
Admin wallet locally = anvil account #0. The runbook covers real deployments: docs/runbook.md.
Checks
pnpm verify # typecheck + all tests + forge test + copy check + React Bits token check
pnpm --filter @stampd/contracts test:fuzz # 10,000 fuzz runs
cd packages/contracts && uvx --from slither-analyzer slither . --config-file slither.config.json
cd apps/scraper && uv run pytest
pnpm --filter @stampd/web test:e2e # Playwright (Edge), against `pnpm dev:local` + the web app
pnpm --filter @stampd/ai live-check # real DeepSeek run on 6 known posts (needs DEEPSEEK_API_KEY in .env)
What the tests prove:
- Contracts: 86 tests including fuzz and invariants:
k never decreases, solvency, supply = collateral − reserve.
- Core: the TypeScript FPMM matches 2,000 Solidity quote vectors to the unit.
- API: 26 contract tests against real Postgres (PGlite).
- Worker: the indexer against anvil (DB matches chain, backfill rebuilds identical rows, reorg rewind, reconciliation alert), plus an end-to-end create → trade → resolve YES/NO/INVALID → finalize → redeem.
- Scraper: kill switch, locked pool, pinned-post ordering.
- E2E: every page renders; the trade panel's quote equals
MarketHub.quoteBuy; reduced motion.
Status against the development plan
Built and tested locally: Phases 0–6 of the development plan, and the code side of Phase 7 (Slither clean of high/medium, CSP, rate limits, key separation in the deploy script).
Needs you or outside parties before launch:
- Testnet deploy (1.8): needs a funded deployer key. The public RPCs are reachable from this machine with WARP on (your ISP's DNS blocks
rpc.*.chain.robinhood.com otherwise).
- Accounts and keys: Postgres and Redis (on Railway), Turnstile, Pinata. Done: DeepSeek key (in the gitignored
.env), two twscrape accounts (V14).
- Launch gates: the external contract review, the multisig (V5), the legal opinion (V11 — the Terms and Privacy pages are marked as drafts), the outage drills on the deployed stack, and the launch KOL list.
