FAQ
Common questions about the Dynoyard gateway.
What is Dynoyard?
A drop-in OpenAI-compatible gateway for frontier OSS models (Kimi,
MiMo, DeepSeek, Qwen, GLM). You point any OpenAI SDK at
https://<your-org>.dynoyard.app/v1, keep your code, and pay
per-token from a prepaid balance.
How is this different from calling the provider directly?
One endpoint and one key for the whole catalog, a dedicated per-org
subdomain, per-key access controls + spend caps, a native usage
dashboard, and inline cost reporting (usage.cost) on every response.
Switch models per request via the body’s model field — no new key,
no new endpoint.
Which models can I use?
Whatever is enabled on your org — call GET /v1/models for the
canonical list, or see the pricing page. New models appear
in that list automatically; no client change needed. See
supported models.
Can I rotate or revoke a key?
Yes. Dashboard → Keys → rotate/revoke. The old key starts
returning 401 within seconds. See
authentication.
Can I restrict a key to certain models?
Yes — set allowed models per key (e.g. a staging key that only
sees mimo-v2-5-pro). A key can never widen the org’s enabled set,
only narrow it.
Do you support function-calling / tools?
Yes, on models flagged with the tools capability in /v1/models.
Send OpenAI-shaped tools[] + tool_choice, get OpenAI-shaped
tool_calls back. See OpenAI compatibility.
Do you support streaming?
Yes — set stream: true. Server-Sent Events in OpenAI’s standard
shape; the final event carries usage + usage.cost. Thinking-mode
models are best consumed with stream: true (long reasoning passes
can exceed non-streaming timeouts).
Can I see my usage?
Dashboard → Usage: per-request token counts, latency, status,
model, and key, last 90 days. Programmatic totals via
GET /v1/management/credits. See usage metering.
What gets stored?
Usage metadata (token counts, latency, status, model id, key id) — yes. Prompt and completion bodies — no, not retained for training or analytics. Two narrow exceptions: an opt-in response cache holds a copy transiently (per-org, short TTL) and the first 1KB of an upstream error body is captured for debugging on failed requests. See privacy & security.
Can I get an SLA?
Volume customers (above $5k/mo) can get a custom SLA + dedicated upstream capacity. Email hello@dynoyard.app.
How do I get help?
Email hello@dynoyard.app — responses within a few business hours (EU / GCC). Status + incidents: status.dynoyard.app.