Skip to content

PRD.md

Latest commit

 

History

History
98 lines (80 loc) · 3 KB

PRD.md

File metadata and controls

98 lines (80 loc) · 3 KB

Product Requirements Document: Pump.fun Launch Radar + Rug-Risk Scanner

Overview

Real-time detection and risk analysis system for new Pump.fun token launches on Solana.

Goals

  1. Detect Pump.fun token creation events in near real-time via webhooks
  2. Generate explainable, heuristic-based rug-risk scores for new tokens
  3. Provide queryable API for launch data and risk reports
  4. Maintain production-grade reliability (idempotency, error handling, observability)

Non-Goals

  • Not a financial advice platform
  • Not a trading bot or execution engine
  • Not a front-end application (API only)
  • Not a comprehensive security audit (heuristics only)
  • Not multi-chain (Solana only for MVP)

Scope

In Scope

  • Webhook ingestion (POST /webhooks/helius)
  • Pump.fun create instruction detection
  • Risk scoring based on:
    • Mint authority presence
    • Freeze authority presence
    • Holder concentration (top holder %, top 5%)
    • Metadata existence
  • REST API endpoints:
    • GET /launches (list recent)
    • GET /tokens/:mint (detail + risk)
    • GET /healthz
  • SQLite persistence with Prisma
  • Async processing queue
  • Idempotent webhook handling
  • CI/CD with quality gates

Out of Scope (Future)

  • Real-time WebSocket updates
  • Historical backfill
  • Advanced ML-based risk models
  • Social sentiment analysis
  • On-chain interaction tracing
  • Multi-provider webhook support
  • Alerting/notification system

User Stories

As a researcher, I want to see all recent Pump.fun launches so I can analyze token creation patterns.

As a trader, I want to check the rug-risk score of a specific token before investing.

As a developer, I want to integrate launch data into my application via REST API.

Acceptance Criteria

Webhook Handler

  • ✓ Accepts POST /webhooks/helius with raw transaction arrays
  • ✓ Validates Authorization header (bearer token)
  • ✓ Returns 200 within 100ms (async processing)
  • ✓ Deduplicates by transaction signature
  • ✓ Validates payload shape with Zod

Detection

  • ✓ Identifies Pump.fun create instructions (program ID + discriminator match)
  • ✓ Extracts mint address from instruction accounts
  • ✓ Stores launch event in LaunchEvent table

Risk Scoring

  • ✓ Computes score 0-100 based on heuristics
  • ✓ Assigns label: LOW / MED / HIGH
  • ✓ Provides human-readable reasons array
  • ✓ Fetches on-chain data (mint info, holder accounts, metadata)
  • ✓ Handles RPC failures gracefully

API

  • ✓ GET /launches returns sorted list with pagination
  • ✓ GET /tokens/:mint returns launch + risk report
  • ✓ Returns "pending" status if risk not yet computed
  • ✓ GET /healthz checks database connectivity

Quality

  • ✓ All tests pass
  • ✓ TypeScript strict mode
  • ✓ ESLint + Prettier
  • ✓ CI workflow runs on PR
  • ✓ Structured logging with pino
  • ✓ No secrets in code

Success Metrics

  • Webhook latency p95 < 100ms
  • Risk computation time p95 < 5 seconds
  • Zero duplicate launch events
  • API uptime > 99%
  • Test coverage > 70%

Timeline

MVP: 1-2 days (this implementation)