Example Index
Examples demonstrate extension points; they are not the only way to enable a capability. Complete the source installation, then follow each example's README for dependencies and configuration. Examples usually run on separate ports—do not substitute the standard Gateway address.
| What you want to explore | Start here |
|---|---|
| Visual conversation, on-demand capture, and observation | X-Omni: Qwen Omni by default, with continuous audiovisual input for MiniCPM-o. |
| WebRTC audio transport | WebRTC: an optional client-to-Gateway transport, without changing the upstream model protocol. |
| Vehicle events, device controls, and a business Agent | Smart Cockpit. |
| A hardware voice client | AI Passport. |
| External long-term memory | VoiceMem. |
| Hosted API-backed memory | Memcode. |
| An external knowledge system | LightRAG. |
| Customer service and human operators | Customer Service: retail and airline scenarios. |
| Frontend and full-system evaluation | Benchmark. |
Choose an extension point
- A voice model: implement a Realtime Provider.
- An action-taking Agent: implement a Backend Adapter.
- A client: use the Gateway Client Protocol.
- Chat tools: configure MCP or OpenAPI.
- Knowledge or memory: implement a Knowledge Provider or Memory Provider.
Business prompts, device actions, and specialized tools stay in the examples. See the extension overview for the boundaries.