v1.2.29

1.2.29 (2026-10-21)

Full Changelog: v1.2.28...v1.2.29

Features

• training-sdk: let a serverless FiretitanServiceClient sample a base model or another run's saved checkpoints without creating a model and without using a concurrent-training-run slot, via standalone sampler routes
• training-sdk: make FiretitanSamplingClient picklable across worker processes (connection state, shared-storage R3 binding, fixed concurrency limit), and add credential-free sampler handles with get_sampler_handle() / FiretitanSamplingClient.from_sampler_handle(handle, api_key=None)
• training-sdk: add stable TITO parser reason codes and payload-free renderer, finish, token-count, and budget diagnostics while preserving parser discard behavior

Bug Fixes

• dependencies: require certifi 2026.10.19 or newer to reject compromised eMudhra certificate chains
• dependencies: update locked aiohttp to 3.14.4 and its required yarl dependency to 1.25.1 to address the SDK security audit findings
• training-sdk: bind a training client's shared-storage (R3) store to a sampler only when it targets that service's own training session, and reject empty or cross-account model_path values before any request
• training-sdk: reject unsuccessful inference streams before committing TITO turns, retain upstream request/status and retry diagnostics, and report known trajectory termination intent in closing/closed errors