generated: '2026-07-20'
method: searched
source: https://www.nixtla.io/docs
authentication:
  style: bearer-api-key
  header: 'Authorization: Bearer <NIXTLA_API_KEY>'
  env_var: NIXTLA_API_KEY
  base_url_env: NIXTLA_BASE_URL
  docs: https://www.nixtla.io/docs/setup/setting_up_your_api_key
  ref: authentication/nixtla-authentication.yml
base_url:
  production: https://api.nixtla.io
  preview: https://api-preview.nixtla.io
idempotency:
  supported: false
  note: >-
    No idempotency-key header or parameter is documented; forecasting/anomaly
    calls are computed request/response and carry no client-supplied idempotency
    key. Fine-tune create returns a server-generated model id.
pagination:
  supported: false
  note: >-
    Batch request/response over the series payload; list operations
    (/v2/finetuned_models) return the full set without cursor/offset paging.
versioning:
  style: uri-path
  current: v2
  ref: lifecycle/nixtla-lifecycle.yml
request_headers:
  - name: Authorization
    purpose: Bearer API key
  - name: nixtla-model
    purpose: Select the model to run (e.g. timegpt-1, timegpt-1-long-horizon, azureai).
  - name: nixtla-client-version
    purpose: SDK version reported by the client.
compression:
  note: The SDK applies zstd compression automatically for request bodies over 1MB (SDK >= 0.6.5).
serialization:
  request: application/json
  note: SDK uses orjson for fast JSON serialization (SDK >= 0.6.0).
error_envelope:
  shape: 'FastAPI HTTPValidationError ({"detail": [{loc, msg, type}]})'
  content_type: application/json
  ref: errors/nixtla-problem-types.yml
rate_limits:
  note: >-
    API-call and usage limits are governed per subscription plan; a `usage`
    method (SDK >= 0.6.4) reports API-call consumption. No public numeric
    rate-limit table is published.