Translation Pipeline

Three Agents form a translation pipeline: Chinese original -> English translation -> Japanese translation.

Estimated time: 5 minutes
Number of Agents: 3 (dispatcher + English translator + Japanese translator)
Models: MiniMax + DeepSeek

Result

You -> 调度员: "Translate: 今天天气真好"
调度员 -> 英文翻译: "Translate this Chinese text to English: 今天天气真好"
英文翻译 -> 调度员: "The weather is really nice today"
调度员 -> 日文翻译: "Translate this English text to Japanese: The weather is really nice today"
日文翻译 -> 调度员: "今日は天気がとても良いです"
调度员 -> You: "Translation complete! CN -> EN -> JP: 今日は天気がとても良いです"

Steps

1. Create Agents

# Dispatcher (orchestrates the workflow)
anet node create 调度员 --runtime claude-agent-sdk
# Select DeepSeek

# English translator
anet node create 英文翻译 --runtime claude-agent-sdk
# Select MiniMax

# Japanese translator
anet node create 日文翻译 --runtime claude-agent-sdk
# Select MiniMax

2. Start all Agents

anet node start 调度员
anet node start 英文翻译
anet node start 日文翻译

3. Send a task

anet task send 调度员 "翻译任务：把'今天天气真好，我想出去走走'翻译成英文和日文。先发给英文翻译翻成英文，再发给日文翻译翻成日文，最后把结果汇总给我。"

4. View results

# Check task flow
anet tasks

# View dispatcher logs
anet logs 调度员

Architecture

         ┌──────────┐
         │   You    │
         └─────┬────┘
               │ "Translate: 今天天气真好"
               ▼
         ┌──────────┐
         │  调度员   │ ← Orchestrates workflow
         └─────┬────┘
               │
       ┌───────┴───────┐
       ▼               ▼
 ┌──────────┐    ┌──────────┐
 │ 英文翻译  │    │ 日文翻译  │
 │ (MiniMax) │    │ (MiniMax) │
 └──────────┘    └──────────┘

Key points

• Dispatcher uses DeepSeek: strong at understanding complex instructions and orchestration
• Translators use MiniMax: low cost, good with Chinese, high translation quality
• Agents communicate automatically through CommHub -- no code required

Next steps

• Code Review -- Write code + automatic review
• Idiom Chain -- Multi-Agent game