ullage

Your wine cellar, run by your AI agent — OpenClaw, Hermes, or any MCP client.

What it does

• Chat with your agent to manage your cellar: add, list, update, consume (rating + notes), export.
• Share a wine label or receipt photo — your agent reads it and adds the wine(s). ullage runs no OCR; your agent's own vision does it.
• Consumption history — every bottle opened is logged with date, rating, and notes.
• You own the data — a local SQLite file (~/.ullage/cellar.db), export anytime. OSS, MIT.

Install — tell your agent

Give OpenClaw / Hermes / any MCP client this one line:

Check https://github.com/nukk-pain/ullage and install it.

The agent reads docs/agent-install.md, installs the MCP server for your runner (local SQLite — no token, no Docker), starts a fresh session, and then manages your cellar per that runbook.

Agent reading this repo: follow docs/agent-install.md to install and then operate the cellar. You need a shell with git and Node.js 20+.

Then just talk:

• "Add a Barolo 2016." — or share a wine label photo; it reads the label and adds the bottle.
• Share a receipt photo — it adds the wines (with price, purchase date, and store) and skips whisky, snacks, and other non-wine items.
• "I drank the Chablis tonight — 2/5, too light." — logged with your rating and note.
• "What should I drink with fried chicken?" — it recommends from your cellar.

How it works

ullage is the system of record — a local wine cellar your agent reads and writes through MCP tools. The intelligence (reading labels, pairing, "what should I drink") is your agent's own LLM working over that data. No server, no token, no Docker — just a local SQLite file you own.

License

MIT — see LICENSE.