JevCodeJev technical solutions and best practices
中ENJAKODEFRESPT

JevCode / SDKs & integration

SDKs & integration

Official client SDKs, the HTTP API, and the skill for AI coding agents.

Source: docs.typesafe.ai/sdksdkapi
SDK language bindings

Three ways to integrate

Route Best for Notes
Python SDK Backend services, data pipelines, batch jobs Sync and async clients, typed inputs, automatic retries
JavaScript SDK Node services, full-stack apps Answer types inferred from your questions
HTTP API Other languages, lightweight integration You handle retries and rate limits yourself

If an AI coding agent is writing the integration, install the TypeSafe agent skill first so it knows the exact request and response shapes instead of guessing.

Shared conventions

All SDKs follow the same conventions:

  • Endpoint: POST https://api.typesafe.ai/v1/systemone
  • Auth: read from the TYPESAFE_API_KEY environment variable; no key in code
  • Default model: jev-latest (resolves to the latest stable release)
  • Retries: backoff by default, honouring the retry-after header when present

Version requirements

  • Python SDK: package typesafe-sdk
  • JavaScript SDK: package @typesafe-ai/sdk, Node.js 20 or newer

The JS SDK ships ESM, CommonJS, and TypeScript declaration files.

Calling the HTTP API directly

Without an SDK you own two things the SDKs already handle:

Rate-limit retries. Exceeding 250,000 tokens/second or 1,200 requests/minute returns 429 Too Many Requests. The response may carry a retry-after header, and you should back off accordingly.

Response parsing. The return value is an answer object keyed by question name, with a different field shape per question type. See the API reference.