@volter/twin-xai
v1.0.0
Published
xAI API twin (Protocol 3): chat completions and responses (streamed and not; tools, JSON output, reasoning) on the World's scenario or the kernel wire's labeled stub (the twin runs no model); Console API keys. Built on @volter/world-core.
Readme
@volter/twin-xai
Protocol 3 xAI inference for Twenty workspace agents and the Volter World Grok chat redirect. The measured callers are in journeys/demand.json.
Serves POST /v1/chat/completions and POST /v1/responses, JSON and SSE, function calls/results and structured output. The kernel OpenAI wire renders scripted scenarios or labelled deterministic stubs. No model or live web/X search runs. The Responses output includes the current model's encrypted reasoning placeholder where documented.
POST /_twin/app-credentials issues or recovers the World's application key (XAI_API_KEY). POST /_twin/api-keys {name} stands in for creating an additional Console key. Keys are drawn from World secrets and stored as hashes; requests use Bearer authorization.
The catalogue is the vendor documentation snapshot, with the spec's latest sample identifier accepted without resolving it to a particular model. Stored response retrieval, previous-response continuation, deferred completion and the remaining spec operations are outside measured scope and use the declared gap. See spec/SOURCE.md for provenance and STANDING.md for generated standing.
