generated: '2026-07-17'
method: searched
source: https://docs.adaptionlabs.ai/introduction/getting-started/index.md
docs: https://docs.adaptionlabs.ai/
api: Adaptive Data API
base_url: https://api.prod.adaptionlabs.ai
summary: >-
  Cross-cutting runtime semantics for the Adaption (Adaptive Data) API, derived
  from the developer docs and the official Python SDK. The API exposes one
  lifecycle — ingest, adapt, evaluate, export — over dataset resources.
auth:
  style: bearer
  header: Authorization
  detail: HTTP Bearer API key (pt_live_ prefix). See authentication/adaption-labs-authentication.yml
versioning:
  style: uri-path
  current: v1
  detail: All endpoints are namespaced under /api/v1 (e.g. /api/v1/datasets).
async_jobs:
  model: poll
  detail: >-
    Ingestion (Hugging Face / Kaggle import) and adaptation runs execute
    asynchronously server-side. Clients poll dataset status and block on
    completion before the next lifecycle stage.
  status_operation: GET /api/v1/datasets/{dataset_id}/status
  sdk_helper: datasets.wait_for_completion()
uploads:
  style: presigned
  detail: >-
    Large file ingestion uses a two-step presigned upload — initiate returns a
    presigned URL, the client PUTs the file, then completes the upload.
  operations:
  - POST /api/v1/datasets/upload/initiate
  - POST /api/v1/datasets/upload/complete
  supported_formats: [csv, json, jsonl, parquet]
media_type:
  request: application/json
  response: application/json
idempotency:
  supported: false
  detail: No idempotency-key header is documented for the Adaptive Data API as of this pass.
pagination:
  style: cursor
  detail: >-
    listDatasets (GET /api/v1/datasets) uses cursor pagination. Pass cursor
    (from the previous response next_cursor) and limit (max 100, default 20);
    the response carries next_cursor (null when no more results). Also supports
    filtering (created_after, created_before, q, status) and sort/sort_direction.
  request_params: [cursor, limit]
  response_fields: [next_cursor]
error_envelope:
  shape: 'object { code, level, message }'
  detail: >-
    Structured failures surface on error_data with a stable code (e.g. E0100),
    a level (error|warning), and a message. See errors/adaption-labs-problem-types.yml.
metering:
  model: credits
  detail: >-
    Runs reserve and consume credits. runDataset with estimate=true returns
    estimatedCreditsConsumed and estimatedMinutes without starting a run;
    multimodal (image) rows bill at 10 credits per 100 output rows.
rate_limits:
  detail: No public rate-limit headers or quota policy documented; usage is metered in credits.
cross_links:
  authentication: authentication/adaption-labs-authentication.yml
  packages: packages/adaption-labs-packages.yml
  changelog: changelog/adaption-labs-changelog.yml