from flask import Flask, request, jsonify
from livekit import api
import os
from datetime import datetime

app = Flask(__name__)
# Put your credentials in environment variables
app.config['LIVEKIT_API_KEY'] = os.getenv('LIVEKIT_API_KEY')
app.config['LIVEKIT_API_SECRET'] = os.getenv('LIVEKIT_API_SECRET')
app.config['LIVEKIT_HOST'] = os.getenv('LIVEKIT_HOST', '<wss://your-livekit-host.com>')

@app.route('/api/agent-token', methods=['POST'])
def generate_agent_token():
    data = request.json
    room_name = data.get('room_name')
    if not room_name:
        return jsonify({'error': 'Missing room_name'}), 400
    agent_identity = f"agent_{datetime.now().timestamp()}"
    
    token_builder = (
        api.AccessToken()
        .with_identity(agent_identity)
        .with_name(agent_identity)
        .with_grants(api.VideoGrants(
            room_join=True, room=room_name, can_publish=True, can_subscribe=True, can_publish_data=True
        ))
        .with_kind('agent')
    )
    
    token = token_builder.to_jwt()
    return jsonify({
        'token': token,
        'room_name': room_name,
        'agent_identity': agent_identity
    })

if __name__ == '__main__':
    app.run(debug=True, port=5000)

{"room_name": "YOUR_ROOM_NAME"}

RoomAgentDispatch

agent = Agent(
    instructions="You are a storyteller. Spin engaging, interactive stories using the information the user provides, and make the experience fun and dynamic."
)

from livekit.agents import AgentSession, Agent
from livekit.plugins import openai

agent = Agent(instructions="You are a storyteller...")
session = AgentSession(
    llm=openai.realtime.RealtimeModel.with_azure(
        azure_deployment="your-gpt-4o-mini-realtime-deployment",
        azure_endpoint="<wss://your-endpoint.openai.azure.com/>",
        api_key="your-api-key",
        api_version="2024-10-01-preview"
    ),
)
await session.start(agent=agent, room=ctx.room)