name: senseaudio-realtime-agent description: Implement and debug SenseAudio realtime Agent lifecycle APIs: list agents, invoke session, query status, and leave session (/v1/realtime/agents, /invoke, /status, /leave). Use this whenever user asks to start or manage SenseAudio conversational agents, continue dialogues, or handle room/conv/token lifecycle errors. metadata: openclaw: requires: env: - SENSEAUDIO_API_KEY primaryEnv: SENSEAUDIO_API_KEY homepage: https://senseaudio.cn compatibility: required_credentials: - name: SENSEAUDIO_API_KEY description: API key from https://senseaudio.cn/platform/api-key env_var: SENSEAUDIO_API_KEY

SenseAudio Realtime Agent

Use this skill for SenseAudio Agent session lifecycle integration.

Read First

• references/agent.md

Workflow

1. Discover agent:

• List available agents and pick agent_id.

2. Start or continue dialogue:

• new_dialogue=true for new session.
• new_dialogue=false with conv_id for continuity.

3. Persist runtime credentials:

• Store conv_id and room_id in your application state (database or session store), never in client-side code or logs.
• Tokens returned by /invoke are short-lived — treat them like passwords: do not log, do not embed in URLs, and discard after the session ends via /leave.
• Rotate by calling /invoke again with the same conv_id; do not reuse expired tokens.

4. Operate session:

• Query room status when needed.
• Leave session explicitly when finished.

5. Handle failures:

• Distinguish quota/auth/not-found vs parameter errors.