v2.29.0

2.29.0 (2026-10-13)

Features

• accumulate streaming continuation usage metadata and expose initialPromptTokenCount (dbc961f)
• Add continuation_token to the Interactions interaction.status_update event. (a93776c)
• Add a CONTINUATION_REQUIRED Interaction status (754f71f)
• add billable token counts fields (839ffc2)
• Add startOffset/endOffset to Transcription and interimTranscriptTimestampEnabled to RealtimeInputConfig (bdbf156)
• Add usage in InteractionStatusUpdate events (8892aea)
• automatically consume the continuation token (98637d1)
• cli: update Speakeasy generator to v2.948.0, bump CLI to v0.4.2, and apply CLI overlay fixes (22d53ec)

Bug Fixes

• clear intermediate CONTINUATION finishReason in streaming continuation (22706af)
• do not record incomplete continuation stream turns in curated chat history (81c3c49)
• gemini-api-cli: remove unsupported 2.5 and 1.6 models from Interactions model options and CLI examples (1002720)
• js: do not set initial prompt token count for requests that do not use automatic consumption of continuation token (0885fb9)
• js: fix streaming continuation edge cases (stale token, AbortError, initialPromptTokenCount) (f3d10a6)
• Raise default HTTP timeouts so long-running requests aren't killed client-side (10e850c)
• resume streaming generation from the latest checkpoint continuation token on mid-stream errors or cutoffs (625ac2c)
• Update EnvironmentConfig.env schema to map keys to EnvVar or string. (8478e12)