| # Manual Setup Guide
|
|
|
| No AI help needed. Do it yourself.
|
|
|
| ## Steps
|
|
|
| 1. **Open the `brain/` folder** inside the downloaded `relic` folder
|
|
|
| 2. **Copy template files to active files** (templates have `.template` in the name — they're blank forms):
|
| - `SOUL.template.md` → rename copy to `SOUL.md`
|
| - `USER.template.md` → rename copy to `USER.md`
|
| - `MEMORY.template.md` → rename copy to `MEMORY.md`
|
|
|
| 3. **Open `SOUL.md`** in any text editor (Notepad, TextEdit, VS Code — anything works) and fill in at least the "Core Mission" section
|
|
|
| 4. **Open `USER.md`** and fill in your preferences
|
|
|
| 5. **(Mac/Linux only) Lock the files** to prevent accidental changes:
|
| ```
|
| chmod 444 SOUL.md USER.md
|
| ```
|
| This locks the files against accidental modification. Unlock first if you need to edit later. Windows users skip this.
|
|
|
| 6. **Point your AI tool** to `AGENT.md` (English) or `AGENT.zh-CN.md` (Chinese) — it'll take over from there
|
|
|
| ## The Seven Files
|
|
|
| | File | Purpose |
|
| |------|---------|
|
| | `SOUL.md` | AI's soul — name, personality, mission, style |
|
| | `USER.md` | Your info — preferred name, preferences, habits |
|
| | `MEMORY.md` | Long-term memory — preferences, decisions, experiences (extracted knowledge, not raw chat) |
|
| | `SKILLS/` | Skills — learned capabilities, one folder per skill |
|
| | `PROJECTS/` | Projects — documentation for things worked on |
|
| | `SESSIONS/` | Conversation logs — full raw chat transcripts with the AI |
|
| | `ARCHIVE/` | Archive — old memories moved here when too long, never deleted |
|
|
|
| ## Connect Your AI Tool
|
|
|
| | AI Tool | How to Connect |
|
| |---------|---------------|
|
| | OpenClaw | See `examples/integrations/openclaw.md` |
|
| | Hermes | See `examples/integrations/hermes.md` |
|
| | Claude Code | See `examples/integrations/generic.md` |
|
| | Cursor | See `examples/integrations/generic.md` |
|
| | Other | See `examples/integrations/generic.md` |
|
|
|
| All tools follow the same pattern: tell the AI to read `AGENT.md`, it handles the rest.
|
|
|
| Having issues? See `docs/TROUBLESHOOTING.md`.
|
|
|
