@cloudflare/think@0.5.3

cloudflare/agents@cloudflare/think@0.5.3May 2, 2026by github-actions[bot]

AI Summary

This patch release (0.5.3) exposes the stable AI SDK's `streamText` call settings on Think's `TurnConfig`, specifically adding `timeout` and `maxRetries` options that can be configured via the `beforeTurn` callback to tune generation behavior on a per-turn basis.

Key Highlights

  • Exposed AI SDK `streamText` settings on Think `TurnConfig`
  • Added `timeout` configuration for per-turn generation behavior
  • Added `maxRetries` configuration for per-turn generation behavior
  • Settings can be tuned via the `beforeTurn` callback

New Features

  • Exposed `timeout` setting from AI SDK's `streamText` on `TurnConfig`
  • Exposed `maxRetries` setting from AI SDK's `streamText` on `TurnConfig`
  • Both settings can be configured per-turn through `beforeTurn` callback

Full Release Notes

### Patch Changes

-   [#1447](https://github.com/cloudflare/agents/pull/1447) [`c7998b2`](https://github.com/cloudflare/agents/commit/c7998b29e54d0a865550c322c76f0ce5d68181ab) Thanks [@threepointone](https://github.com/threepointone)! - Expose stable AI SDK `streamText` call settings on Think `TurnConfig`, including `timeout` and `maxRetries`, so `beforeTurn` can tune generation behavior per turn.