[ download_resume.pdf ]
← back to projects
mathew@dfw:~/projects/trenchwatch
guest@portfolio:~$cat trenchwatch/README.md

[WIP] TrenchWatch started: 2024-11
tech: Telethon, SQLite, Rugcheck API, Dexscreener API, systemd, Ubuntu 24.04

What It Does

TrenchWatch monitors Solana token launches via Telegram channel feeds and the Dexscreener API. For each new token it:

  1. Ingests the contract address from Telegram signal sources via Telethon
  2. Runs a Phase 5A pre-filter to discard obvious noise: missing Rugcheck data, zero liquidity, or immediate rug flags
  3. Classifies surviving tokens across four behavioral buckets:
    • Runner — price action and volume sustaining an upward trajectory
    • Slow Cook — gradual accumulation pattern, not spiking yet
    • Rug — liquidity drain detected, holders dumping
    • Dead — flatlined, zero volume, abandoned

Architecture

  • Telethon listener — async Python process subscribing to Telegram channel events
  • SQLite store — lightweight local database tracking token state, timestamps, classification history, and outcome labels
  • Dexscreener API — price, volume, and liquidity polling on a configurable interval per-token
  • Rugcheck API — contract risk score at intake; high-risk tokens bypassed immediately
  • systemd unit — runs as a persistent service on this VPS, separate from the portfolio site

Shadow Mode

TrenchWatch is currently running in shadow mode: it classifies tokens and logs outcomes to the database but does not place any trades or send actionable signals. The shadow run exists to collect labeled outcome data and validate classifier accuracy before any live signal output is enabled.

Phase 5A pre-filter logic is active. Phase 5B (confidence thresholds based on shadow data) is in development.

type help to see available commands
guest@portfolio:~$