What problem does it solve? Navigating the Nexi voice assistant codebase is difficult because logic is spread across many engine modules for wake detection, ASR, intent routing, tools, and UI. This Skill provides an architecture map so you can orient quickly and trace any feature end to end. ## Core Features & Use Cases - Layer-by-layer codebase map: Documents where wake/audio, ASR, intent routing, model selection, tools, brain, agency, forge, Claude Code driver, and UI bridge code lives under engine/. - Provider boundary conventions: Clarifies that Groq handles ASR and intent while Gemini handles chat and vision, and warns against confusing Groq with xAI's Grok. - Tool registration rules: Explains that a tool must appear in _TOOLS, ALLOWED_INTENTS, and TOOL_INTENTS in engine/tool_registry.py or it silently fails to route. - Use Case: When adding a new desktop automation capability, use this Skill to learn that you must register it in tool_registry.py across three structures and route it through groq_intent_router_v2.py. ## Quick Start Ask the assistant to explain where the intent routing logic lives in the Nexi codebase and how a voice command flows from wake detection to a spoken response.
