JasonLiu0826/ackem

Ack — A local‑first AI companion for Windows

What it is – Ackem (Autonomous Companion Keeping Emotional Memory) is a Windows‑only desktop application built with Electron. It lets you hook any OpenAI‑compatible language model (cloud services like OpenAI or DeepSeek, or local servers such as Ollama/LM Studio) and provides a persistent “friend” that can chat, remember past conversations, track a simple emotional/relationship state, and even appear as a tray icon or optional desktop pet.

Key capabilities

• Chat – Sends prompts to the configured LLM and displays replies in a chat window.
• Structured memory – Every conversation is stored locally, indexed with embeddings, searchable, and can be visualised as timelines/knowledge graphs. You can also import plain‑text (.txt) or markdown (.md) files as long‑term context.
• Companion state – The app maintains trust, mood, relationship stage, and a diary. Personality presets and proactive messages let the “companion” feel more alive.
• Desktop presence – System‑tray icon, optional geometric orb/LIVE2D pet, and optional voice STT/TTS.
• Extensions – An Extension Center lets developers add plugins (e.g., WeChat bridge, game‑mode integration, custom skills).
• Local‑first data – All chats, memories, API keys and settings live in a data/ folder next to Ackem.exe. The release zip contains no user data and does not ship telemetry.

How to get started (end‑user)

1. Download the green release zip from the GitHub or Gitee releases page.
2. Extract the zip to a folder on an SSD (do not run from inside the zip).
3. Run Ackem.exe (or the provided batch file). The first launch extracts an embedding model and takes ~10‑30 s.
4. Accept the privacy/compliance screen.
5. Open Settings → Model & API and enter the base URL, model ID and API key for the LLM you want to use.
6. Start chatting; import any .txt/.md files via the Memory view to give the companion long‑term knowledge.

Developer notes

• The project is an Electron app with a Vite‑based renderer. Development requires Node 20+ on Windows.
• Core scripts: npm run dev (run the app in development mode), npm run build (compile), npm run dist:green (produce the portable release), npm run dist:setup (NSIS installer).
• Architecture is split into seven logical systems (Brain, Heart, Mouth, Neural, Extension, Time, etc.) documented in docs/developer/architecture/.
• The codebase includes TypeScript type‑checking, unit tests, and a SQLite‑backed data layer.

Privacy & data handling

• The official release does not ship any user data. After first run, everything lives in the local data/ directory.
• API keys are stored only in the local settings file; no telemetry is sent to an Ackem server.
• The repository contains a detailed privacy policy (docs/privacy-and-data.md).

Status – Actively tested by a single maintainer. Crashes and lag are reported as rare but may still occur; users are encouraged to open issues for unexpected behaviour.

License – AGPL‑3.0. Personal, research and open‑source integration use is permitted. Commercial or closed‑source SaaS use requires a separate commercial license (contact provided in the README).

---

All information above is taken directly from the repository’s README and linked documentation.