DEVELOPER DOCUMENTATION

Tester onboarding

A small, inspectable first run for invited testers.

1. Open the deployed Router and select the intended gateway/network. Check gateway health and the routable model catalog.
2. Connect your own wallet and create a dedicated key in API Keys. Copy it once and keep it in your server environment. The gateway must enable key creation and inference.
3. For paid requests, make sure your wallet has active stake and available Compute Credits. Only advertised, enabled :free routes use the separate sponsor allowance.
4. Start with public, non-sensitive text and a 64-token cap in the playground. Try only exact model IDs advertised by your gateway.
5. Inspect the provider, tokens, cost, timing and outcome in Activity. Export metadata for feedback, keeping credentials and private content out of reports.
6. Try privacy redaction or rejection. Use the latency benchmark for a bounded direct comparison once both provider credentials are available.

Compatibility and availability

The API currently supports text Chat Completions. Provider families such as Anthropic, Gemini, Grok or GLM are usable only when the selected gateway advertises that exact model and supports its route. Catalog descriptions do not enable images, tools or other unsupported dimensions.

What is ready to evaluate

Model discovery and price charts, key management, metered routing, optional privacy processing, request receipts and latency instrumentation are implemented. Benchmark tooling is available, but live performance numbers require an actual run. Deferred batches are implemented behind encryption/inference configuration and enforceable provider price caps; current live adapters cannot fill them yet.

A tester invitation should include the deployed URL, selected network, usable model IDs and how credits are funded. Do not promise a launch date, universal savings or negligible overhead before those are measured. Creating this flow does not provision a tester’s key or notify anyone automatically.