ADUC

Agents Dress Up Committee

Local-first familiars for your workflow. Ship code, not your data.

PG-13 by default. Playful when you want it, focused when you need it.

Download Quickstart

Playful, mod-friendly, and yours. MIT licensed.

No Cloud, No Telemetry

Runs locally. Your files, your rules. Bring your own model or CLI agent.

Modular Familiars

Each familiar lives in a folder: agent.md, personality.md, state.json, and avatar/*.png.

File-based CLI Bridge

Simple JSON handoff in <temp>/ADUC/conversation.json. Any CLI can roleplay by appending replies.

Emotion-Tagged Replies

End responses with <emotion: focus>. UI swaps avatars automatically.

Coach + Crush Vibes

Your familiar keeps you on task, celebrates wins, and, with your opt-in, adds a playful tone. Work first; flirting later.

How it works

  1. Pick a familiar

    Edit files in familiars/<name>/ to change role, vibe, and visuals.

  2. Merge the prompt

    ADUC combines agent.md + personality.md + current state.json.emotion (+ optional memory, lore).

  3. Bridge to your CLI

    When the UI writes a user turn, your CLI appends a reply with an emotion tag.

// Minimal watcher (pseudocode)
pending = turns.filter(t => t.role == 'user' && t.status == 'pending')
for (u of pending) append({ role: 'cli', in_reply_to: u.id,
  text: 'Show me the code.\n<emotion: focus>' })

Quickstart

1) Install (Windows)

py -3 -m pip install --user -r requirements.txt

2) Launch

.\run_aduc_simple.bat 8080

3) Optional: CLI bridge

py tools/cli_bridge_watcher.py --familiar lumi

Tip: Responses should end with a state line like <emotion: calm>.

Download

Grab the code and run locally.

Download .zip View on GitHub

You can update these links later. Set data-github on the <body> or edit in HTML.

Safety & Boundaries