Skip to main content
This guide uses the current CogFoundry API structure. The earlier ShengSuanYun guide used legacy /batch/v1 paths; new integrations should use the /loom/v1 operations generated from the OpenAPI document in this repository.

​
1. Check the service

​
2. Discover available models and official templates

Use the returned template identifier. Do not infer availability or field names from an example.

​
3. Inspect the selected template

The schema returned by the active environment defines the accepted field keys, value types, required inputs, and available choices.

​
4. Validate, estimate, and confirm

Choose the row or workbook operation that matches your input:
Do not create a run before the user has reviewed the current estimate. If the payload changes, repeat validation and precheck and obtain a new confirmation.

​
5. Run and preserve the returned ID

After confirmation, call runRows or runWorkbook and preserve the returned runId exactly. Use a new client request ID for every newly confirmed execution; reuse one only for an identical retry after an ambiguous failure.

​
6. Retrieve progress and results

Generated artifact download URLs may be temporary. Retrieve current result metadata instead of storing signed URLs in long-lived logs.
Use the generated endpoint pages for exact request bodies, parameters, response schemas, and error models. This quickstart explains the safe sequence rather than duplicating the OpenAPI contract.