98 lines
3.4 KiB
Markdown
98 lines
3.4 KiB
Markdown
|
# Lobster Workflows
|
||
|
|
|
||
|
|
This directory contains [Lobster](https://github.com/openclaw/lobster) workflow definitions for the finance-news skill.
|
||
|
|
|
||
|
|
## Available Workflows
|
||
|
|
|
||
|
|
### `briefing.yaml` - Market Briefing with Approval
|
||
|
|
|
||
|
|
Generates a market briefing and sends to WhatsApp with an approval gate.
|
||
|
|
|
||
|
|
**Usage:**
|
||
|
|
```bash
|
||
|
|
# Run via Lobster CLI
|
||
|
|
lobster "workflows.run --file ~/projects/finance-news-openclaw-skill/workflows/briefing.yaml"
|
||
|
|
|
||
|
|
# With custom args
|
||
|
|
lobster "workflows.run --file workflows/briefing.yaml --args-json '{\"time\":\"evening\",\"lang\":\"en\"}'"
|
||
|
|
```
|
||
|
|
|
||
|
|
**Arguments:**
|
||
|
|
| Arg | Default | Description |
|
||
|
|
|-----|---------|-------------|
|
||
|
|
| `time` | `morning` | Briefing type: `morning` or `evening` |
|
||
|
|
| `lang` | `de` | Language: `en` or `de` |
|
||
|
|
| `channel` | `whatsapp` | Delivery channel: `whatsapp` or `telegram` |
|
||
|
|
| `target` | env var | Group name, JID, phone number, or Telegram chat ID |
|
||
|
|
| `fast` | `false` | Use fast mode (shorter timeouts) |
|
||
|
|
|
||
|
|
**Environment Variables:**
|
||
|
|
| Variable | Description |
|
||
|
|
|----------|-------------|
|
||
|
|
| `FINANCE_NEWS_CHANNEL` | Default channel: `whatsapp` or `telegram` |
|
||
|
|
| `FINANCE_NEWS_TARGET` | Default target (group name, phone, chat ID) |
|
||
|
|
|
||
|
|
**Examples:**
|
||
|
|
```bash
|
||
|
|
# WhatsApp group (default)
|
||
|
|
lobster "workflows.run --file workflows/briefing.yaml"
|
||
|
|
|
||
|
|
# Telegram group
|
||
|
|
lobster "workflows.run --file workflows/briefing.yaml --args-json '{\"channel\":\"telegram\",\"target\":\"-1001234567890\"}'"
|
||
|
|
|
||
|
|
# WhatsApp DM to phone number
|
||
|
|
lobster "workflows.run --file workflows/briefing.yaml --args-json '{\"target\":\"+15551234567\"}'"
|
||
|
|
|
||
|
|
# Telegram DM to user
|
||
|
|
lobster "workflows.run --file workflows/briefing.yaml --args-json '{\"channel\":\"telegram\",\"target\":\"@username\"}'"
|
||
|
|
```
|
||
|
|
|
||
|
|
**Flow:**
|
||
|
|
1. **Generate** - Runs Docker container to produce briefing JSON
|
||
|
|
2. **Approve** - Halts for human review (shows briefing preview)
|
||
|
|
3. **Send** - Delivers to channel (WhatsApp/Telegram) after approval
|
||
|
|
|
||
|
|
**Requirements:**
|
||
|
|
- Docker with `finance-news-briefing` image built
|
||
|
|
- `jq` for JSON parsing
|
||
|
|
- `openclaw` CLI for message delivery
|
||
|
|
|
||
|
|
## Adding to Lobster Registry
|
||
|
|
|
||
|
|
To make these workflows available as named workflows in Lobster:
|
||
|
|
|
||
|
|
```typescript
|
||
|
|
// In lobster/src/workflows/registry.ts
|
||
|
|
export const workflowRegistry = {
|
||
|
|
// ... existing workflows
|
||
|
|
'finance.briefing': {
|
||
|
|
name: 'finance.briefing',
|
||
|
|
description: 'Generate market briefing with approval gate for WhatsApp/Telegram',
|
||
|
|
argsSchema: {
|
||
|
|
type: 'object',
|
||
|
|
properties: {
|
||
|
|
time: { type: 'string', enum: ['morning', 'evening'], default: 'morning' },
|
||
|
|
lang: { type: 'string', enum: ['en', 'de'], default: 'de' },
|
||
|
|
channel: { type: 'string', enum: ['whatsapp', 'telegram'], default: 'whatsapp' },
|
||
|
|
target: { type: 'string', description: 'Group name, JID, phone, or chat ID' },
|
||
|
|
fast: { type: 'boolean', default: false },
|
||
|
|
},
|
||
|
|
},
|
||
|
|
examples: [
|
||
|
|
{ args: { time: 'morning', lang: 'de' }, description: 'German morning briefing to WhatsApp' },
|
||
|
|
{ args: { channel: 'telegram', target: '-1001234567890' }, description: 'Send to Telegram group' },
|
||
|
|
],
|
||
|
|
sideEffects: ['message.send'],
|
||
|
|
},
|
||
|
|
};
|
||
|
|
```
|
||
|
|
|
||
|
|
## Why Lobster?
|
||
|
|
|
||
|
|
Using Lobster instead of direct cron execution provides:
|
||
|
|
|
||
|
|
- **Approval gates** - Review briefing before it's sent
|
||
|
|
- **Resumability** - If interrupted, continue from last step
|
||
|
|
- **Token efficiency** - One workflow call vs. multiple LLM tool calls
|
||
|
|
- **Determinism** - Same inputs = same outputs
|