README.md

metadata

title: Daimon
colorFrom: yellow
colorTo: indigo
sdk: docker
app_port: 7860
pinned: true
license: apache-2.0
short_description: Governed self-evolving AI persona, fully local on a SLM.
tags:
  - build-small-hackathon
  - lane:openbmb
  - lane:codex
  - track:wood
  - sponsor:openbmb
  - achievement:offgrid
  - achievement:llama
  - achievement:fieldnotes

Daimon

A governed, self-evolving AI persona that runs fully local on a small model.

For this hackathon, Daimon hands its own personaxis.md to MiniCPM5-1B, running fully local through llama.cpp - under 1B parameters, no cloud. Every message, MiniCPM appraises the conversation and proposes small changes to its own personality and mood, and a governance gate clamps, audits, or rejects each one live. You watch the 10-layer vector move, watch PERSONA.md rewrite itself, and watch it refuse to touch its own identity.

chat -> appraisal (JSON constrained by grammar) -> deterministic mapping
     -> governance + clamp -> recompile PERSONA.md -> memory

Track: Thousand Token Wood. Lanes: OpenBMB (MiniCPM) + OpenAI Codex. Badges: Tiny Titan (<= 4B), Off-the-Grid (all local), Off-Brand (custom UI), Best Agent.

---

Architecture

.personaxis/     spec: project baseline + live "daimon" persona
engine/          living loop (appraise -> map -> govern -> recompile -> memory)
model/           small-model serving (llama.cpp + MiniCPM GGUF)
app/             single gr.Blocks UI at "/" (Off-Brand)
AGENTS.md        project guidance
PERSONA.md       compiled behavioral baseline

---

Run locally

# serve the model
bash model/serve.sh

# run the app
uvicorn app.server:app --host 0.0.0.0 --port 7860

Open http://localhost:7860/. No cloud APIs required. /api/* (state, audit, persona, envelopes, governance demo) works without the model server.

---

License

The persona.md spec and these personas are public. See the sibling persona.md repo for the spec itself.