Run a private AI chat locally with your own API key and no cloud accounts.
See how the assistant's remembered facts change or expire over time.
Save a good chat answer as a reusable document you can reference later.
Let the assistant read, edit, and search files in a local project folder.
| lnshrivas/minimachat | 100/rutgers-pbl-dining-2015 | a15n/a15n_old | |
|---|---|---|---|
| Stars | 0 | — | — |
| Language | HTML | HTML | HTML |
| Last pushed | — | 2015-12-01 | 2016-06-18 |
| Maintenance | — | Dormant | Dormant |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 1/5 | 1/5 |
| Audience | vibe coder | general | general |
Figures from each repo's GitHub metadata at analysis time.
Windows-only today, requires Python 3.10+, PowerShell, and your own API key for an OpenAI-compatible model.
minimachat, also called minima, is a local chat application that talks directly to an AI model of your choice while running entirely on your own computer. Its main idea is that it keeps its memory of past facts honest over time. If it remembers a deadline, and that deadline passes, the app visibly strikes it through instead of quietly continuing to treat it as still true. Beyond that, the app lets you drag a timeline slider to scrub back and see exactly what the assistant believed at an earlier point, almost like rewinding its memory. When you tell it to save a good answer, it turns that answer into a standalone document, called a book, that you can reference again later from any other chat, along with a link back to the exact moment it was written. The assistant can also act on your files directly: reading, writing, editing, and searching your local folders, running commands, and searching the web, all without sending anything to an outside company beyond your chosen AI model provider. You bring your own API key and can connect it to any AI service that follows the common OpenAI style API format. Right now the project only runs on Windows, since its local server is written in PowerShell, though the author has flagged a Python based server as the most wanted community contribution to bring it to macOS and Linux. Setting it up involves cloning the repository, installing some Python dependencies for memory and web search, and starting the included server script, after which you open a page in your browser and enter your API details. The project is released under the MIT license, a permissive license that allows free use, modification, and distribution as long as the copyright notice is kept. Some bundled third party components carry their own separate licenses.
A local AI chat app whose memory of facts ages honestly over time, letting you rewind to see what it believed earlier and save answers as reusable documents.
Mainly HTML. The stack also includes HTML, JavaScript, PowerShell.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice, some bundled parts have their own licenses.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
double-check against the repo, no cap.