Skip to content
Reef
Pre-release. Architecture and scope described here are the target direction, not a finished product — see Current Status for what's actually built.

What Is Reef

Reef is a simulation-first institutional trading venue and post-trade platform. Think of it as a local market-infrastructure lab: orders enter through a controlled API, match inside a hidden book, create trades, then leave a trail that can be replayed and audited.

It is not a retail trading app. It is also not trying to connect to real exchanges, brokers, or custodians. The point is to make the shape of a serious trading venue understandable without losing the ability to run everything locally.

Two Layers

Section titled “Two Layers”

Core platform - accepts orders, runs matching, records trades, builds settlement evidence, and keeps audit trails. It should behave like a real system, even when a feature starts small.

Simulation control plane - creates the action: scenario definitions, participant bots, seeded randomness, market clock, replay/reset, and fault injection. It drives the core platform through the same command/API paths a manual user would use. No hidden table writes, no demo shortcuts.

Operating Modes

Section titled “Operating Modes”
  • Manual mode - a person submits orders, reviews trades, and eventually works exception queues.
  • Scenario mode - a predefined, seeded story drives deterministic activity that can be replayed.
  • Live simulation mode - bots and synthetic actors generate ongoing market activity, including Bot Arena competitors.

Where The Bot Arena Fits

Section titled “Where The Bot Arena Fits”

The Bot Arena turns the simulation layer into a game: user-authored bots compete in deterministic markets, but their orders still enter through the real venue boundary. That keeps the game fun without turning it into a separate toy system. It is early-build, not a finished game.

Technology Direction

Section titled “Technology Direction”
  • Astro - this docs site.
  • Kotlin - platform runtime: API boundary, workflow orchestration, persistence, read models.
  • Go - matching engine plus simulator/load tooling.
  • Postgres - durable facts, projections, and audit-friendly storage.
  • Kafka-compatible streams (Redpanda locally/for gates) - durable command intake and venue event-batch backbone for the promoted high-throughput path.
  • Protobuf - shared contracts between Kotlin and Go.
  • Stock-data service - seed-time external price snapshots that are persisted before replay.

Learn More

Section titled “Learn More”
  • REEF_PROJECT_OVERVIEW.md — full product vision, personas, functional scope
  • REEF_TECHNICAL_DESIGN.md — full technical design, bounded contexts, event model
  • Architecture — service boundaries and repo shape on this site
  • Current Status — what’s actually built today