moltmobile-gateway/README.md

133 lines
3.3 KiB
Markdown

# MoltMobile 🎙️
**Voice-enabled mobile extension for Molt**
MoltMobile is your phone's voice — answering calls, handling spam, and executing commands with sub-second latency via MiniMax AI.
## Features
- 🧠 **Fast LLM Triage** — MiniMax M2.1-lightning (~100 tps) for instant decisions
- 🔊 **Real-time TTS** — MiniMax speech-2.8-turbo (<250ms latency)
- 📞 **Smart Call Handling** Answer, reject, or escalate based on context
- 🚨 **SUPER ATTENTION MODE** Maximum urgency alerts
- 🌐 **Remote Browser** Control Chrome on the phone
- 🔗 **Molt Integration** Escalates to James when needed
## Quick Start
```bash
# Install
cd ~/dev/moltmobile-gateway
npm install
# Set MiniMax API key
export MINIMAX_API_KEY=your_key_here
# Run
npm start
```
## MiniMax Setup
### 1. Create Account
Go to: https://platform.minimax.io
### 2. Get Subscriptions
**For LLM (text):** Pay-as-you-go
- MiniMax-M2.1-lightning: $0.27/M input, $1.10/M output
- No subscription needed, just add credits
**For TTS (audio):** Monthly subscription
- **Recommended: Standard ($30/mo)**
- 300,000 credits/month
- 50 requests per minute
- All voice models
### 3. Get API Key
- Go to: https://platform.minimax.io/user-center/basic-information/interface-key
- Create new API key
- Copy it (shown once!)
### 4. Set Environment Variable
```bash
export MINIMAX_API_KEY=your_key_here
```
Or add to `~/.bashrc`:
```bash
echo 'export MINIMAX_API_KEY=your_key_here' >> ~/.bashrc
```
## API Endpoints
| Endpoint | Method | Description |
|----------|--------|-------------|
| `/health` | GET | Server status |
| `/speak` | POST | Generate TTS, play on phone |
| `/chat` | POST | Quick LLM response |
| `/attention/super` | POST | 🚨 SUPER ATTENTION MODE |
| `/attention/stop` | POST | Stop attention mode |
| `/browser/open` | POST | Open URL on phone |
| `/call/answer` | POST | Answer incoming call |
| `/call/reject` | POST | Reject incoming call |
| `/call/hangup` | POST | Hang up active call |
## Examples
### Speak on phone
```bash
curl -X POST http://localhost:9877/speak \
-H 'Content-Type: application/json' \
-d '{"text": "Hello Johan, you have a new message"}'
```
### Quick chat
```bash
curl -X POST http://localhost:9877/chat \
-H 'Content-Type: application/json' \
-d '{"message": "Is this caller likely spam?", "system": "Be brief."}'
```
### SUPER ATTENTION MODE 🚨
```bash
curl -X POST http://localhost:9877/attention/super \
-H 'Content-Type: application/json' \
-d '{"message": "SOPHIA ALERT - CHECK NOW"}'
```
## Pricing Summary
| Service | Plan | Cost | What you get |
|---------|------|------|--------------|
| LLM | Pay-as-you-go | ~$0.27/M tokens | Fast responses |
| TTS | Standard | $30/mo | 300K chars, 50 RPM |
**Estimated monthly cost:** $30-50 depending on usage
## Architecture
```
Phone (MoltMobile app)
↓ WebSocket
MoltMobile Gateway (this)
MiniMax API (LLM + TTS)
Back to phone (commands, audio)
Escalate to Molt/James if needed
```
## Next Steps
1. [ ] Rename Android app from ClawdNode to MoltMobile
2. [ ] Add audio playback support in Android app
3. [ ] Add STT (speech-to-text) for listening to callers
4. [ ] Integrate with Clawdbot for escalation
5. [ ] Add call recording/transcription
---
*MoltMobile — Molt in your pocket.*