Skip to content

maintainer.md

Latest commit

 

History

History
119 lines (108 loc) · 7.03 KB

maintainer.md

File metadata and controls

119 lines (108 loc) · 7.03 KB
Raw

Maintainer Notes

TODO — Wave uplift (deadline: next wave entry ≈ 2026-07-23)

Uplift plan

  • 1. Deploy reputation contract — mainnet. Run the existing scripts/deploy-oracle-mainnet.ts script, record the contract ID in the repo, and wire submitToOracle (packages/publisher/src/batch.ts) for real: build the Soroban invoke tx, sign with the publisher key, submit. The publisher tick must publish real outcomes on-chain BEFORE entry. This is the categorical D3 move (6 → 8-9). Deliberately deferred — deployment stays testnet-only for now. Everything downstream of it (publisher tick, on-chain reads, leaderboard reconciliation) is wired and verified end-to-end against the existing testnet contract instead; only the mainnet deploy + contract-id swap is outstanding.
  • 2. App reads scores from chain. Reconcile/verify the leaderboard against contract state and surface on-chain tx hashes in the UI. Closes the oracle loop in both directions — no mock-adjacent gap. Done against testnet. New lib/oracle/read.ts is a read-only Soroban client (get_score_for_corridor, get_corridor_aggregate, list_anchors — pure simulateTransaction, no signing/funds needed; verified against the live deployed contract). /api/reputation/leaderboard no longer returns hardcoded STUB_METRICS — it computes real scorecards from outcome_log and, when a corridor filter is given, attaches a live on-chain score per anchor for reconciliation. app/anchors/[id]/page.tsx was actually surfacing the wrong tx hash (the off-chain settlement payment, not an oracle submission) — fixed to source the real oracle_tx_hash written by the publisher (see fix-order item 2 below).
  • 3. Publish @stellarintel/mcp + @stellarintel/publisher to npm with a runnable example wired to the REAL contract ID; make examples/consumer-contract invoke the deployed oracle. "Primitive others depend on" = only observed path to S. Both packages build + npm pack --dry-run clean; packages/mcp now actually registers its two tools for real (see fix-order item 1) instead of shipping an empty server. examples/consumer-contract has a runnable read-oracle.mjs verified live against the real testnet contract, plus a README covering both that and the full Rust deploy/invoke path. Did not run the actual npm publish — needs the maintainer's npm login/2FA and is hard to fully reverse once published.

Fix order before publish

  • 1. Wire real tools into packages/mcp/src/server.ts from lib/mcp/offramp.ts — with LIVE rates, not routing-table constants. intel.offramp.quote/intel.offramp.prepare are now registered for real (was an empty tool list), sourcing lib/mcp/offramp.ts directly via a tsc-alias-rewritten build so the published dist/ stays self-contained outside this monorepo. getQuote no longer computes netReceived off the static ANCHOR_ROUTING rate — it calls fetchCorridorRates for the routed anchor's actual current price (SEP-38 firm quote, falling back to SEP-24/SEP-6 fee-adjusted live FX) and throws RATE_UNAVAILABLE when that anchor can't currently be quoted, rather than returning a stale number. Needed two small Node16 compat fixes along the way in lib/stellar/sep1.ts and lib/stellar/anchors.ts, and fixed the bin field noted in item 4 below.
  • 2. Wire submitToOracle to the deployed contract (blocks on contract deploy — item 1 of the uplift plan). Add pg + @stellar/stellar-sdk to packages/publisher dependencies. Done against the testnet deployment (CCZ54NTEOVL2DKWCGJA5XHTHOGRDS7JHFKYWEC6QH2IMZLYNM3FBFKDG); mainnet wiring still pending item 1 of the uplift plan. Found and fixed a real bug on top of this: fetchPendingOutcomes/ markPublished in batch.ts queried a table called reputation_outcomes, which the app never creates — its real table is outcome_log. packages/publisher's own e2e test seeds through the app's real store, so this would have failed the moment it ran against an actual database. Renamed to outcome_log and added the published_at/oracle_tx_hash columns it needs (migration lib/reputation/migrations/005_oracle_publish.sql). Also made packages/publisher's index.ts safe to import (it previously ran its CLI main(), including process.exit(1), unconditionally at module load) and wired /api/publisher/tick — previously a hardcoded no-op stub — to actually call runBatch against the testnet oracle.
  • 3. Fix both package tsconfigs (module: "Node16", exclude tests from build), verify npm run build emits dist/, and check npm pack --dry-run shows the expected files. Both packages now build and pack clean. Along the way: fixed a pre-existing @types/node auto-discovery gap (TS 6 stopped finding it in this workspace layout — added explicit "types": ["node"]), and worked around @stellar/stellar-sdk shipping ESM-only types against publisher's CommonJS build via a dynamic import() in batch.ts. Also added a local vitest.config.mts for packages/publisher (it was silently inheriting the root app's config, pointed at a setup file that doesn't exist there) and gitignored packages/*/dist.
  • 4. Add README + LICENSE + repository + "publishConfig": {"access": "public"} to each package. Also added "license": "MIT" to each package.json (implied by shipping a LICENSE file — avoids the npm-publish "no license field" warning) and refreshed docs/ORACLE_SPEC.md to point at the real testnet contract instead of describing it generically. Noted but did not fix: packages/mcp/package.json's bin key is the full scoped name ("@stellarintel/mcp"), which contains / — likely invalid as an npm bin command name; worth checking before publish. Fixed (see fix-order item 1): bin is now the bare dist/... path string, letting npm derive the unscoped command name.

11. Anchor Fleet Status

  • Monthly recheck complete for the latest survey snapshot.

Latest documented snapshot: 92 directory-tagged domains -> 32 reachable stellar.toml files -> 9 transfer-capable / 23 issuer-only; 60 unreachable or unconfirmed.

The 23 issuer-only domains (advertise an asset/issuer but no SEP-6/SEP-24 transfer rails, so they back no corridor) are enumerated with reasons in docs/anchors/exclusions.md.

Source: scripts/anchor-survey.snapshot.json, generated 2026-06-25T23:08:26.806Z from https://api.stellar.expert/explorer/public/directory?tag[]=anchor&limit=200.

Refresh cadence: re-run node scripts/anchor-survey.mjs --json monthly, update these counts, and keep the recheck checkbox aligned with the current snapshot.