Skip to content

MOSH-1979: Sync RL API OpenAPI spec with latest shaping protos to incorporate training checkpoint saving/resuming#227

Merged
sbassam merged 2 commits intomainfrom
sbassam/update-rl-specs
Mar 17, 2026
Merged

MOSH-1979: Sync RL API OpenAPI spec with latest shaping protos to incorporate training checkpoint saving/resuming#227
sbassam merged 2 commits intomainfrom
sbassam/update-rl-specs

Conversation

@sbassam
Copy link
Contributor

@sbassam sbassam commented Mar 17, 2026

Summary

Syncs the OpenAPI spec with api.swagger.yaml (together-shaping) for the RL save/resume training checkpoint feature.

Changes

Session create (Start)

  • RL.StartTrainingSessionRequest: Replaced checkpoint_id with resume_from_checkpoint_id (optional string) so the spec matches the backend. Description: "Checkpoint ID to resume from."

Session response

  • RL.TrainingSession: Added:
    • training_checkpoints: Array of RL.TrainingCheckpoint (list of saved training checkpoints for the session).
    • resume_from_checkpoint_id: Optional string (checkpoint this session was resumed from).

New endpoints

  • POST /rl/training-sessions/{session_id}/operations/training-checkpoint
    • createTrainingCheckpoint: Submits a save-training-checkpoint operation (adapter + optimizer + step to object storage). Returns RL.TrainingCheckpointOperation.
  • GET /rl/training-sessions/{session_id}/operations/training-checkpoint/{operation_id}
    • getTrainingCheckpointOperation: Returns status and result of a save-training-checkpoint operation.

New schemas

  • RL.TrainingCheckpoint: id, step, created_at (saved training checkpoint item).
  • RL.TrainingCheckpointResult: checkpoint_id (used for resume via Start).
  • RL.TrainingCheckpointOperation: id, status, output, error (same shape as inference checkpoint operation).

@sbassam sbassam requested review from VProv and khaykingleb March 17, 2026 03:53
@github-actions
Copy link

github-actions bot commented Mar 17, 2026

✱ Stainless preview builds

This PR will update the togetherai SDKs with the following commit messages.

go

chore(internal): regenerate SDK with no functional changes

openapi

feat(api): add training checkpoint operations/models to RL, update session types

python

chore(internal): regenerate SDK with no functional changes

terraform

chore(internal): regenerate SDK with no functional changes

typescript

chore(internal): regenerate SDK with no functional changes
togetherai-openapi studio · code

Your SDK build had at least one "note" diagnostic.
generate ✅

togetherai-python studio · code

Your SDK build had at least one "note" diagnostic.
generate ✅build ✅lint ✅test ⏭️

togetherai-go studio · code

Your SDK build had at least one "note" diagnostic.
generate ✅build ⏭️lint ✅test ✅

go get github.com/stainless-sdks/togetherai-go@0c14b26cf18b273ab587e9e0a8f6ee49be8c9381
togetherai-typescript studio · code

generate ✅build ⏳lint ⏳test ⏳

togetherai-terraform studio · code

Your SDK build had at least one "note" diagnostic.
generate ✅lint ✅test ✅

⏳ These are partial results; builds are still running.


This comment is auto-generated by GitHub Actions and is automatically kept up to date as you push.
If you push custom code to the preview branch, re-run this workflow to update the comment.
Last updated: 2026-03-17 21:31:40 UTC

Co-authored-by: Gleb Khaykin <khaykingleb@gmail.com>
@sbassam sbassam merged commit e72f5c9 into main Mar 17, 2026
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants