Skip to content

quick-start.md

Latest commit

 

History

History
110 lines (87 loc) · 3.45 KB

quick-start.md

File metadata and controls

110 lines (87 loc) · 3.45 KB
sidebar_position 3
title Quick Start
description Build and run your first Conversational AI agent in Python.

Quick Start

This guide walks you through building a voice agent using the cascading flow (ASR → LLM → TTS) with both sync and async clients.

Sync Example

This complete script creates an agent with Aliyun for the LLM, MiniMax for TTS, and Fengming for STT:

from shengwang_agent import AgentClient, Area
from shengwang_agent.agentkit import Agent
from shengwang_agent.agentkit.vendors import AliyunLLM, MiniMaxTTS, FengmingSTT

# 1. Create a client with app credentials
client = AgentClient(
    area=Area.CN,
    app_id='your-app-id',
    app_certificate='your-app-certificate',
)

# 2. Build an agent with vendor configuration
agent = (
    Agent(name='support-assistant', instructions='你是一个智能语音助手。')
    .with_llm(AliyunLLM(api_key='your-aliyun-key', model='qwen-max'))
    .with_tts(MiniMaxTTS(key='your-minimax-key', voice_id='your-voice-id'))
    .with_stt(FengmingSTT(language='zh-CN'))
)

# 3. Create and start a session
session = agent.create_session(
    client,
    channel='support-room-123',
    agent_uid='1',
    remote_uids=['100'],
)
agent_id = session.start()
print(f'Agent started with ID: {agent_id}')

# 4. Interact with the agent
session.say('你好！有什么可以帮助你的？')

# 5. Stop the session when done
session.stop()
print('Agent stopped.')

Async Example

For async applications, use AsyncAgentClient for the client. All session methods become coroutines that require await:

import asyncio
from shengwang_agent import AsyncAgentClient, Area
from shengwang_agent.agentkit import Agent
from shengwang_agent.agentkit.vendors import AliyunLLM, MiniMaxTTS, FengmingSTT

async def main():
    # 1. Create an async client
    client = AsyncAgentClient(
        area=Area.CN,
        app_id='your-app-id',
        app_certificate='your-app-certificate',
    )

    # 2. Build an agent (same as sync — Agent is client-agnostic)
    agent = (
        Agent(name='support-assistant', instructions='你是一个智能语音助手。')
        .with_llm(AliyunLLM(api_key='your-aliyun-key', model='qwen-max'))
        .with_tts(MiniMaxTTS(key='your-minimax-key', voice_id='your-voice-id'))
        .with_stt(FengmingSTT(language='zh-CN'))
    )

    # 3. Create a session — works with both sync and async clients
    session = agent.create_session(
        client,
        channel='support-room-123',
        agent_uid='1',
        remote_uids=['100'],
    )

    # 4. All session methods are coroutines — use await
    agent_id = await session.start()
    print(f'Agent started with ID: {agent_id}')

    await session.say('你好！有什么可以帮助你的？')
    await session.stop()
    print('Agent stopped.')

asyncio.run(main())

What Happens Under the Hood

  1. The Agent builder collects your vendor configuration into a properties object
  2. session.start() generates an RTC token (using the client's app_id and app_certificate), then calls the API to start the agent
  3. The agent connects to the specified channel and begins listening for audio from the remote UIDs
  4. session.say() sends text to be spoken by the agent's TTS
  5. session.stop() gracefully shuts down the agent

Next Steps