Update docs to reflect current architecture

- SPEC.md: Add new source files, update config location, document
  conversation catch-up feature, fix message flow description
- customize/SKILL.md: Fix file references (was Python, now TypeScript),
  update launchd service name

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
gavrielc
2026-01-31 19:36:22 +02:00
parent 0c08e8a034
commit 423d45c52e
2 changed files with 58 additions and 40 deletions

View File

@@ -18,11 +18,12 @@ This skill helps users add capabilities or modify behavior. Use AskUserQuestion
| File | Purpose |
|------|---------|
| `src/config.py` | Assistant name, trigger pattern, settings |
| `src/router.py` | Message routing, polling, agent invocation |
| `src/scheduler_worker.py` | Scheduled task execution |
| `src/commands.py` | Command handlers |
| `.mcp.json` | MCP server configuration |
| `src/config.ts` | Assistant name, trigger pattern, directories |
| `src/index.ts` | Message routing, WhatsApp connection, agent invocation |
| `src/db.ts` | Database initialization and queries |
| `src/types.ts` | TypeScript interfaces |
| `src/auth.ts` | Standalone WhatsApp authentication script |
| `.mcp.json` | MCP server configuration (reference) |
| `groups/CLAUDE.md` | Global memory/persona |
## Common Customization Patterns
@@ -36,9 +37,9 @@ Questions to ask:
- Should messages from this channel go to existing groups or new ones?
Implementation pattern:
1. Find/add MCP server for the channel to `.mcp.json`
2. Add polling function in `router.py` (similar to `get_new_messages()`)
3. Add to main loop to poll both sources
1. Find/add MCP server for the channel
2. Add connection and message handling in `src/index.ts`
3. Store messages in the database (update `src/db.ts` if needed)
4. Ensure responses route back to correct channel
### Adding a New MCP Integration
@@ -49,8 +50,8 @@ Questions to ask:
- Which groups should have access?
Implementation:
1. Add MCP server config to `.mcp.json`
2. Add tools to `allowed_tools` in `router.py`
1. Add MCP server to the `mcpServers` config in `src/index.ts`
2. Add tools to `allowedTools` array
3. Document in `groups/CLAUDE.md`
### Changing Assistant Behavior
@@ -59,7 +60,7 @@ Questions to ask:
- What aspect? (name, trigger, persona, response style)
- Apply to all groups or specific ones?
Simple changes → edit `src/config.py`
Simple changes → edit `src/config.ts`
Persona changes → edit `groups/CLAUDE.md`
Per-group behavior → edit specific group's `CLAUDE.md`
@@ -71,8 +72,8 @@ Questions to ask:
- Does it need new MCP tools?
Implementation:
1. Add handler function in `src/commands.py`
2. Claude will recognize natural language and call the function
1. Add command handling in `processMessage()` in `src/index.ts`
2. Follow the pattern used for `/clear`
### Changing Deployment
@@ -89,9 +90,10 @@ Implementation:
Always tell the user:
```bash
# Restart to apply changes
launchctl unload ~/Library/LaunchAgents/com.nanoclaw.router.plist
launchctl load ~/Library/LaunchAgents/com.nanoclaw.router.plist
# Rebuild and restart
npm run build
launchctl unload ~/Library/LaunchAgents/com.nanoclaw.plist
launchctl load ~/Library/LaunchAgents/com.nanoclaw.plist
```
## Example Interaction
@@ -100,7 +102,7 @@ User: "Add Telegram as an input channel"
1. Ask: "Should Telegram use the same @Andy trigger, or a different one?"
2. Ask: "Should Telegram messages create separate conversation contexts, or share with WhatsApp groups?"
3. Find Telegram MCP (e.g., telegram-mcp)
4. Add polling for Telegram in router.py
5. Update .mcp.json
3. Find Telegram MCP or library
4. Add connection handling in index.ts
5. Update message storage in db.ts
6. Tell user how to authenticate and test

58
SPEC.md
View File

@@ -88,10 +88,13 @@ nanoclaw/
├── .gitignore
├── src/
── index.ts # Main application (WhatsApp + routing + agent)
── index.ts # Main application (WhatsApp + routing + agent)
│ ├── config.ts # Configuration constants
│ ├── types.ts # TypeScript interfaces
│ ├── db.ts # Database initialization and queries
│ └── auth.ts # Standalone WhatsApp authentication
├── dist/ # Compiled JavaScript (gitignored)
│ └── index.js
├── .claude/
│ └── skills/
@@ -118,7 +121,7 @@ nanoclaw/
│ ├── sessions.json # Active session IDs per group
│ ├── archived_sessions.json # Old sessions after /clear
│ ├── registered_groups.json # Group JID → folder mapping
│ └── router_state.json # Last processed timestamp
│ └── router_state.json # Last processed timestamp + last agent timestamps
├── logs/ # Runtime logs (gitignored)
│ ├── nanoclaw.log # stdout
@@ -132,16 +135,17 @@ nanoclaw/
## Configuration
Configuration is done via environment variables and the CONFIG object in `src/index.ts`:
Configuration constants are in `src/config.ts`:
```typescript
const CONFIG = {
assistantName: process.env.ASSISTANT_NAME || 'Andy',
pollInterval: 2000, // ms
storeDir: './store',
groupsDir: './groups',
dataDir: './data',
};
export const ASSISTANT_NAME = process.env.ASSISTANT_NAME || 'Andy';
export const POLL_INTERVAL = 2000;
export const STORE_DIR = './store';
export const GROUPS_DIR = './groups';
export const DATA_DIR = './data';
export const TRIGGER_PATTERN = new RegExp(`^@${ASSISTANT_NAME}\\b`, 'i');
export const CLEAR_COMMAND = '/clear';
```
### Changing the Assistant Name
@@ -152,9 +156,9 @@ Set the `ASSISTANT_NAME` environment variable:
ASSISTANT_NAME=Bot npm start
```
Or edit the default in `src/index.ts`. This changes:
Or edit the default in `src/config.ts`. This changes:
- The trigger pattern (messages must start with `@YourName`)
- The response prefix (`YourName:`)
- The response prefix (`YourName:` added automatically)
### Placeholder Values in launchd
@@ -248,16 +252,16 @@ When a user sends `/clear` in any group:
└── Is message "/clear"? → Yes: handle specially
6. Router prepares invocation:
├── Load session ID for this group
├── Determine group folder path
└── Strip trigger word from message
6. Router catches up conversation:
├── Fetch all messages since last agent interaction
├── Format with timestamp and sender name
└── Build prompt with full conversation context
7. Router invokes Claude Agent SDK:
├── cwd: groups/{group-name}/
├── prompt: user's message
├── resume: session_id (or undefined)
├── prompt: conversation history + current message
├── resume: session_id (for continuity)
└── mcpServers: gmail, scheduler
@@ -266,10 +270,10 @@ When a user sends `/clear` in any group:
└── Uses tools as needed (search, email, etc.)
9. Router captures result and sends via WhatsApp
9. Router prefixes response with assistant name and sends via WhatsApp
10. Router saves new session ID
10. Router updates last agent timestamp and saves session ID
```
### Trigger Word Matching
@@ -281,6 +285,18 @@ Messages must start with the trigger pattern (default: `@Andy`):
- `What's up?` → ❌ Ignored (no trigger)
- `/clear` → ✅ Special command (no trigger needed)
### Conversation Catch-Up
When a triggered message arrives, the agent receives all messages since its last interaction in that chat. Each message is formatted with timestamp and sender name:
```
[Jan 31 2:32 PM] John: hey everyone, should we do pizza tonight?
[Jan 31 2:33 PM] Sarah: sounds good to me
[Jan 31 2:35 PM] John: @Andy what toppings do you recommend?
```
This allows the agent to understand the conversation context even if it wasn't mentioned in every message.
---
## Commands