Skip to main content
Both files cover SwitchX API and Enterprise API, regardless of the documentation selector. They are generated from the published documentation.

​
Available Files

llms.txt

Concise index of all documentation pages with summaries. Use this for quick context and discovery.

llms-full.txt

Full documentation content in a single file. Use this when you need complete API details.

​
Choose the API context

Tell your AI tool which API you are using. Keep that API’s request fields, billing, job statuses, response format, and webhooks together. Product job results use outputs; SwitchX generation results use output.

​
Product API workflow

Give your AI tool the Quickstart and the guide for your chosen product, then follow this order:
  1. Authenticate: create an Organization API Key and send x-api-key. Keep the same organization and internal team for uploads and jobs.
  2. Discover: use GET /v1/products and GET /v1/products/{product}/models. Build inputs from the selected model’s input_schema; do not invent model IDs or fields.
  3. Prepare and estimate: pass a Beeble URI, authorized CDN URL, public/presigned HTTPS URL, or base64 data URI in each supported media field. Follow the media limits, including the 50 MiB decoded data URI limit and model-specific requirements. Supported audio references also accept these forms. Estimate credits before submission; estimates may store imported files without charging. Keep external URLs readable with unchanged contents through both calls.
  4. Submit: use a unique idempotency_key for each new job. On an uncertain result, retry with the same key and body.
  5. Retrieve: poll GET /v1/product-jobs/{job_id} until success, or use webhooks. Download the fields shown in the product’s completed response.
The request schema and response schema define the common job structure. Product output examples are illustrative, not exhaustive schemas. Product jobs use Beeble Cloud credits; check the Overview for current availability.

​
SwitchX 2.0 Finish

For a higher-resolution version of a completed SwitchX 2.0 product result, follow SwitchX 2.0 Finish. Use product switchx_finish, model switchx-2.0, and only parent_job_id plus target_resolution inside inputs; no upload is needed. The parent must be a public dap_… generation owned by the same key owner, organization, and team. Finish results and legacy /v1/switchx/generations jobs cannot be parents. Supported transitions are 720 → 1080, 720 → 2160, and 1080 → 2160, subject to source-size limits. Estimate the transition before submitting, keep the same body and idempotency key on uncertain retries, and retrieve outputs from the new child job ID, not the parent.

​
OpenAPI Specification

For programmatic API client generation or tool integrations (MCP, Cursor, etc.), the OpenAPI spec is available at: