Gormes

OPEN SOURCE · MIT LICENSE

One Go Binary. Same Hermes Brain.

A static Go binary that talks to your Hermes backend over HTTP. scp it to Termux, Alpine, a fresh VPS — Gormes adds no runtime of its own on top of what Hermes already needs.

Install View Source

1. INSTALL

curl -fsSL https://gormes.ai/install.sh | sh

2. RUN

gormes

Requires Hermes backend at localhost:8642. Install Hermes →

FEATURES

Why a Go layer matters.

Single Static Binary

Zero CGO. ~17 MB. scp it to Termux, Alpine, a fresh VPS — it runs.

Boots Like a Tool

No Python warmup. 16 ms render mailbox keeps the TUI responsive under load.

In-Process Tool Loop

Streamed tool_calls execute against a Go-native registry. No bounce through Python.

Survives Dropped Streams

Route-B reconnect treats SSE drops as a resilience problem, not a happy-path omission.

SHIPPING STATE

What ships now, what doesn't.

  • SHIPPED Phase 1 — Bubble Tea TUI shell.
  • SHIPPED Phase 2.A–C — Tool registry + Telegram adapter + session resume.
  • NEXT Phase 2.B.2+ — Wider gateway (Discord, Slack, more adapters).
  • SHIPPED Phase 3.A–C + 3.D.5 — SQLite + FTS5 lattice, ontological graph, neural recall, USER.md mirror (Gormes-original, no upstream equivalent).
  • NEXT Phase 3.D — Ollama embeddings + semantic fusion.
  • LATER Phase 4 — Brain transplant. Hermes backend becomes optional.
  • LATER Phase 5 — 100% Go. Python tool scripts ported. Hermes-off.