Realtime Voice

Realtime voice is an operator input surface for claw-bot. It starts and stops the audio session; it does not change the perception/action contract.

中文：实时语音只是操作员输入层。语音负责产生 transcript 和交互状态，不绕过 robot_* MCP 工具，也不改变感知、决策、行动的主契约。

claw-bot voice start --environment realworld --json
claw-bot voice status --json
claw-bot voice stop --json

Capture a read-only, redacted acceptance artifact from the current session:

claw-bot voice evidence --json
claw-bot voice evidence --output /path/to/voice-evidence.json --json

The evidence command never starts, stops, or mutates a voice session. It records a SHA-256 digest of the source state, redacts credential-shaped fields and values, and writes requested artifacts with mode 0600.

Voice transcripts should become normal decision-loop work:

1. inspect robot_status
2. request perception when needed
3. issue robot_act
4. record robot_feedback

中文流程：

1. 语音转写进入 Agent。
2. Agent 用 robot_status 获取当前环境和能力。
3. 需要现场状态时调用 robot_perceive 或 robot_subscribe。
4. 用 robot_act 执行动作。
5. 用 robot_feedback 记录结果并继续规划。

Use environmentId to override the active environment. Otherwise, use the active environment returned by robot_status.

中文：只有需要显式切换环境时才传 environmentId；默认使用 robot_status 返回的 active environment。