Deterministic transaction matching with configurable rules, audit trails, and a REST API. Built as a monorepo with TypeScript, Node.js, and PostgreSQL.
Status: Active development (local). Not yet production-ready. See Intentional Boundaries.
Settler compares financial records from two sources (e.g., a payment processor's statement and a bank statement) and produces a deterministic match report identifying:
- Matched records — records that agree within configured tolerances
- Unmatched source records — records in the source with no corresponding match
- Unmatched target records — records in the target with no corresponding match
- Conflicts — records where reference data exists on both sides but values differ beyond tolerance
Results are stored in PostgreSQL with an append-only audit trail. The system supports tenant-isolated data, configurable matching rules, and idempotent ingestion.
| Layer | Technology |
|---|---|
| Runtime | Node.js 22+ / 24+ |
| Language | TypeScript 5.x |
| API framework | Express 5.x |
| Package manager | pnpm 9.x |
| Database | PostgreSQL 15+ via Prisma ORM |
| Ledger (opt.) | TigerBeetle (Docker, optional) |
| Queue (opt.) | Redis via BullMQ / Upstash |
| Frontend | Next.js (App Router) — operator console |
| CLI tooling | TypeScript CLI via tsx |
See packages/api/package.json and package.json for the full dependency tree.
packages/
api/ Express API server — routes, services, middleware
web/ Next.js operator console
cli/ CLI tooling (foundry, replay, verification)
reconciliation-core/ Core matching engine and run result serialization
types/ Shared TypeScript types
adapters/ Source/target adapters
sdk/ Client SDK
support-intake/ Support ticket integration
proofs/ Proofpack generation utilities
edge-ai-core/ ML matching enhancement (optional)
agents/ Workflow agent runner (experimental)
workhorse/ Python background worker
tools/
settler-engine/ Go reconciliation engine (experimental)
reconciliation_intel/ Drift detection, compliance snapshots (experimental)
reconciliation_engine/ Python reconciliation engine
docs/ Documentation
scripts/ Build, CI, and operational automation
prisma/ Prisma schema and migrations
- Node.js 22.x or 24.x (see
.nvmrcor engine field inpackage.json) - pnpm 9.x (enable via
corepack enable) - Docker (for local PostgreSQL, Redis, and optional TigerBeetle)
- PostgreSQL 15+ (via Docker or native install)
git clone https://github.com/settler/settler.git
cd settler
pnpm run bootstrap # creates .env.local, installs deps, validates setup
pnpm tb:start # starts PostgreSQL + Redis + TigerBeetle (Docker)
pnpm dev # http://localhost:3000 (console), http://localhost:4000 (API)For step-by-step setup instructions, see SETUP.md.
The API exposes versioned routes under /api/v1/. Key endpoints include:
POST /api/v1/ingestion— Ingest source/target records (CSV, JSON, webhooks)POST /api/v1/reconciliation/run— Execute a reconciliation runGET /api/v1/reconciliation/runs— List reconciliation runsGET /api/v1/reconciliation/runs/:id— Reconciliation result detailGET /api/v1/audit-trail— Append-only audit log
Routes require authentication and are scoped by tenant ID. See API Reference for full documentation.
The monorepo includes a comprehensive verification pipeline:
pnpm verify # lint, typecheck, build, test, integrity checks
pnpm test # unit tests
pnpm test:e2e # Playwright-based end-to-end testspnpm dev # Start API + web in dev mode (hot reload)
pnpm run demo:seed # Seed 100+ realistic transaction scenarios
pnpm run doctor # Environment diagnostics- SETUP.md — Canonical local development setup
- What Works Today — Currently functional features
- Intentional Boundaries — Known limitations
- Architecture — System design overview
See CONTRIBUTING.md. All pull requests must pass pnpm verify and pnpm run typecheck.
See LICENSE.