generated: '2026-07-18'
method: searched
source: https://www.asterlab.ai/inference
docs: https://www.asterlab.ai/inference
summary: >-
  Aster Inference is an OpenAI-compatible HTTP API. Cross-cutting request/response semantics
  follow the OpenAI Chat Completions contract, so any client that speaks the OpenAI API works
  by setting the base URL to https://api.asterlab.ai/v1 and passing an Aster API key.
conventions:
  base_url: https://api.asterlab.ai/v1
  protocol: HTTPS / JSON
  compatibility: OpenAI API compatible (chat completions, models); works with the official OpenAI
    Python and JS SDKs, LangChain, LiteLLM, Cursor, the Vercel AI SDK, and most agent frameworks.
  authentication:
    style: bearer
    header: 'Authorization: Bearer <ASTER_API_KEY>'
    ref: authentication/asterlab-authentication.yml
  model_selection:
    param: model
    values:
    - gpt-oss-120b
    - gpt-oss-120b-fast
    - GLM 5.2
  data_retention: >-
    Zero data retention by default — prompts and outputs are processed in memory and never stored;
    only token counts are kept for billing.
  idempotency:
    supported: unknown
    notes: No idempotency-key contract is documented on the inference page (beyond any behavior
      inherited from OpenAI-compatible clients).
  pagination:
    supported: unknown
    notes: Not documented; the published surface is the OpenAI-style chat/completions and models
      endpoints.
  error_envelope:
    notes: OpenAI-compatible error envelope expected (error object with type/message), inherited
      from the OpenAI API contract; not independently documented by Aster.
  rate_limit_signaling:
    notes: Not documented on the public inference page.