On-chain systemsFLAGSHIP / A01

SOL Eater

A Go-based trading engine split into monitoring, filtering, execution, position-management, and compounding services.

PROOF / SYSTEM05 SERVICES / DETERMINISTIC EXECUTION
STATUS
IN DEVELOPMENT
ARCHITECTURE
08 STAGES
STACK
05 SYSTEMS
01 / PROBLEM

React to newly created Solana markets under strict latency and risk constraints while keeping network failures, strategy logic, and transaction execution independently recoverable.

02 / ARCHITECTURE

SYSTEM FLOW

  1. 01Solana RPC / WebSocket
  2. 02Market monitor
  3. 03Normalizer
  4. 04Risk + filter engine
  5. 05Execution queue
  6. 06Transaction builder
  7. 07Submitter
  8. 08Position manager
03 / ENGINEERING DECISIONS

WHY IT IS BUILT THIS WAY

01

Why Go?

Predictable concurrency and low-overhead long-running workers fit continuous monitoring and execution services.

02

Why separate monitoring from execution?

A stalled RPC subscription should never block transaction construction or submission.

03

Why runtime configuration?

Filters, position sizing, and exits need to change without rebuilding or redeploying the engine.

04

Why deterministic rules?

Explicit sizing and exit paths make asynchronous execution easier to inspect, test, and recover.

04 / TECHNICAL CHALLENGES
  • Recovering unreliable RPC and WebSocket connections
  • Coordinating concurrent market observations
  • Preventing duplicate transaction submission
  • Maintaining deterministic state across asynchronous workers
  • Keeping strategy configuration separate from transport
05 / WHAT I BUILT
Architecture
Backend service boundaries, queue flow, and failure isolation.
Backend
Go monitoring, filtering, execution, position, and compounding services.
Frontend
TypeScript and React control surfaces separated from the execution runtime.
Infrastructure
Runtime configuration, RPC integration, deployment, and operational telemetry.
Research
Solana transaction mechanics and market-execution behavior.
06 / STACK
GoSolanaRPC / WebSocketTypeScriptReact

Five independent Go services separate observation from execution. Typed errors, explicit queues, and runtime configuration keep strategy changes out of the deployment path.

NEXT PROJECT / A02Fisidi