Roo Code 3.29.3 Release Notes (2025-10-28)

This patch adds reasoning controls for Roo and Z.ai providers, updates Gemini "latest" models, and delivers stability and UI fixes.

QOL Improvements

Toggle time and cost display in the system prompt to reduce model distraction during long runs (thanks jaxnb!) (#8451)
After “Add to Context,” input auto‑focuses and two newlines are inserted for clearer separation, so you can keep typing immediately (#8877)

LiteLLM: Prefer max_output_tokens (with fallback to max_tokens) to prevent 400 errors on Claude Sonnet 4.5 via Vertex (thanks fabb!) (#8455)
Messages typed during context condensing now send automatically once condensing finishes; per‑task queues no longer cross‑drain (thanks JosXa!) (#8478)
Rate limiting now uses a monotonic clock and enforces a hard cap at the configured limit to avoid long lockouts (thanks chrarnoldus, intermarkec!) (#8456)
Restore green tests and TypeScript build for LiteLLM after interface changes; keeps monotonic clock fix and token limit behavior (#8870)
Checkpoint menu popover no longer clips long option text; items remain fully visible (#8867)

Roo provider: Reasoning effort control lets you choose deeper step‑by‑step thinking vs. faster/cheaper responses. See Roo Code Router for details. (#8874)
Z.ai (GLM‑4.5/4.6): “Enable reasoning” toggle to activate Deep Thinking; hidden on unsupported models (thanks BeWater799!). See Z.ai provider. (#8872)
Gemini: Updated model list and “latest” aliases for easier selection (thanks cleacos!). See Gemini provider. (#8486)