Providers
XYVA's five-provider parity goal is Ollama, LM Studio/Bionic, OpenAI, Claude, and Gemini. A provider is not “supported” until its product-specific readiness and released capability have been verified.
Local and cloud boundaries
Ollama and LM Studio are local-provider options: their service runs on your device. Bionic is a separate agent app powered by the LM Studio runtime and can use the same stable lmstudio provider path when its local API service is running. OpenAI, Claude, and Gemini are cloud-provider options: access is configured through approved agent or product settings. Never post keys in URLs, source control, chat, email, screenshots, or public issues.
Readiness states
- Available: product-specific evidence has been published.
- Prepared: a contract or adapter exists, but a full product runtime has not been published or verified.
- Planned: do not treat as usable.
Flow workflow inference is not available. A separate Provider Preview is prepared but unavailable until the npm next package chain and Flow consumer are published. When enabled for an invited account, it may list models and send one explicit neutral request capped at 32 output tokens; it is not a workflow run and requires entitlement, product-specific pairing, and a local grant for the exact provider.
The candidate keeps provider setup and Flow authorization separate. Each of the five providers is denied to Flow by default and must be granted individually by the local agent owner; changing a cloud credential removes that grant. QA Studio configuration is not automatically inherited as Flow consent.