Skip to content

vespid_voice

vespid_voice is the flagship lane in the current repo set: an Android-first voice control system that routes spoken intent through realtime orchestration, confirmation policy, and agent backends.

  • Surface: Android voice product with orchestrator + agent bridge
  • Repo visibility: private implementation repo
  • Current emphasis: user usability and runtime stability over demo theatrics
  • Product angle: reliable spoken control for operator workflows

The live path already covers a meaningful end-to-end loop:

  • Android joins a realtime voice session
  • Volc handles ASR, VAD, interruption, subtitle events, and TTS
  • the orchestrator forwards turns into an agent backend
  • the session can run against OpenClaw or a local Codex backend

That makes this more than a concept page. It is a working product lane with verified device/runtime behavior.

  • human-to-agent interface, not just text chat
  • structured confirmations and risk-aware control
  • realtime orchestration instead of static request/response wrappers
  • operational polish as a product requirement

Push the product from “working lane” toward “consumer-stable lane”:

  1. login/session recovery
  2. voice session recovery across network issues
  3. state sync between Android and orchestrator
  4. more durable end-to-end stability on the main user path