Sakrylle API

English

简体中文

English

简体中文

Appearance

Service-specific Terms ​

Last updated: May 22, 2026

Overview ​

This document supplements the operational and technical aspects of the Sakrylle API and forms part of the Terms of Service. In the event of any conflict between this document and the Terms of Service, this document prevails.

Service availability ​

We maintain the Sakrylle API on a best-effort basis, but make no specific percentage commitment regarding availability, response time, or throughput. Please understand the following:

  • Actual availability depends heavily on the operational status of upstream model providers (Anthropic, OpenAI, etc.)
  • Failures of network carriers, data centers, or upstream providers may cause uncontrollable interruptions or degradations
  • Planned maintenance is normally announced in advance on the platform; emergency maintenance may proceed without prior notice
  • We do not provide service credits or equivalent compensation tied to availability

If availability is critical to your use case, we recommend implementing timeouts, retries, and fallback strategies on the client side, and avoiding hard dependence on a single upstream model.

Data handling ​

Request forwarding ​

  • User-submitted requests (including prompts and context) are forwarded to the corresponding upstream model provider
  • We do not permanently store, cache, or repurpose request or response bodies
  • The forwarding pipeline operates in memory and is released as soon as the request finishes

Logs and billing records ​

To support operations, billing, and risk control, we record the following metadata:

  • Request timestamp
  • Model name and group
  • Token usage (input, output, cache hits, etc.)
  • Response status code and latency
  • Internal identifier of the API key (the full key is not stored)
  • Request source IP (used for risk control and incident investigation)

We do not record the body of requests or responses. Logs are retained for 30 days and are automatically rotated and deleted afterwards.

Upstream data ​

Once a request body is forwarded to an upstream provider, subsequent handling (whether it is used for training, retention period, regional routing, etc.) is governed by that provider's policies. We recommend the following practices:

  • Do not include personal sensitive information, trade secrets, or regulated data in requests
  • Before submitting sensitive data, review the upstream provider's privacy and data protection policies
  • Apply masking, placeholders, or local preprocessing as needed

Model availability ​

  • The model lists returned by the console and the GET /v1/models endpoint are the live source of truth
  • Models may change for reasons including upstream removal, upstream version updates, upstream access policy changes, and compliance requirements
  • We will try to announce significant changes in advance, but in urgent cases changes may be applied immediately
  • We do not guarantee continuous availability of any specific model over any time window

Pricing and billing ​

  • The price of each model follows what is published in the console; the actual price for the same model in different groups is determined by the group's rate multiplier
  • Upstream cost changes, exchange rate changes, or operational needs may trigger price adjustments
  • Adjusted prices apply to new requests from the moment they take effect; already consumed amounts are not retroactively repriced

Fair use ​

  • Service resources are shared across all users; please avoid sustained request rates far above reasonable levels
  • Anomalous call patterns may trigger automated risk controls and result in temporary throttling or manual verification
  • Specific policies for billing, rate limiting, and key management are maintained on the backend; the limits shown in the console are the values currently in effect

API compatibility ​

  • The Sakrylle API offers both an OpenAI-compatible interface and an Anthropic-native interface; available endpoints are documented in the developer docs
  • We strive to maintain backward compatibility for existing interfaces
  • Unavoidable breaking changes will be announced in advance with migration guidance
  • We recommend pinning call patterns and model names in production, and parsing response structures defensively

Security and accounts ​

  • The console supports password sign-in, GitHub OAuth, and TOTP two-factor authentication (2FA); we recommend enabling 2FA for administrator accounts
  • Password recovery requires email verification and a Cloudflare Turnstile check; please complete the flow through the console UI
  • See the Login Terms for detailed account rules

Technical support ​

  • Technical support is provided by email at support@sakrylle.com
  • Total-outage issues are handled with priority; other issues are handled in the order they are received
  • We do not provide official support over public IM, social platforms, or telephone

Contact ​

For questions about these Service-specific Terms, please reach us at: