|
3 | 3 | All notable changes to this project will be documented in this file. |
4 | 4 |
|
5 | 5 |
|
| 6 | +## [v1.33.0] - 2026-03-18 |
| 7 | + |
| 8 | +This release improves file editing reliability, adds session exit keywords, and fixes several issues with sub-sessions and evaluation handling. |
| 9 | + |
| 10 | +## What's New |
| 11 | +- Adds support for "exit", "quit", and ":q" keywords to quit sessions immediately |
| 12 | +- Adds per-eval Docker image override via evals.image property in evaluation configurations |
| 13 | +- Adds run instructions to creator agent prompt for proper agent execution guidance |
| 14 | + |
| 15 | +## Bug Fixes |
| 16 | +- Fixes handling of double-serialized edits argument in edit_file tool when LLMs send JSON strings instead of arrays |
| 17 | +- Fixes sub-session thinking state being incorrectly derived from parent session instead of child agent |
| 18 | +- Fixes --sandbox flag when running in CLI plugin mode |
| 19 | +- Fixes cross-model Gemini function calls by using dummy thought_signature |
| 20 | +- Fixes event timestamps for user messages in SessionFromEvents to prevent duration calculation issues |
| 21 | + |
| 22 | +## Improvements |
| 23 | +- Displays breakdown of failure types in evaluation summary for better debugging |
| 24 | +- Declines elicitations in run --exec --json mode |
| 25 | +- Validates path field consistently in edit file operations |
| 26 | + |
| 27 | +## Technical Changes |
| 28 | +- Removes unused fileWriteTracker from creator package |
| 29 | +- Simplifies UnmarshalJSON implementation for better path validation |
| 30 | +- Updates evaluation image build cache to handle different images per working directory |
| 31 | + |
| 32 | +### Pull Requests |
| 33 | + |
| 34 | +- [#2144](https://github.com/docker/docker-agent/pull/2144) - fix: handle double-serialized edits argument in edit_file tool |
| 35 | +- [#2146](https://github.com/docker/docker-agent/pull/2146) - Better rendering in tmux and ghostty |
| 36 | +- [#2147](https://github.com/docker/docker-agent/pull/2147) - docs: update CHANGELOG.md for v1.32.5 |
| 37 | +- [#2149](https://github.com/docker/docker-agent/pull/2149) - fix: sub-session thinking state derived from child agent, not parent session |
| 38 | +- [#2150](https://github.com/docker/docker-agent/pull/2150) - Display breakdown of types of failures in eval summary |
| 39 | +- [#2151](https://github.com/docker/docker-agent/pull/2151) - Fix --sandbox when running cli plugin mode |
| 40 | +- [#2152](https://github.com/docker/docker-agent/pull/2152) - feat: support "exit" as a keyword to quit the session |
| 41 | +- [#2153](https://github.com/docker/docker-agent/pull/2153) - Add per-eval Docker image override via evals.image property |
| 42 | +- [#2154](https://github.com/docker/docker-agent/pull/2154) - Add run instructions to creator agent prompt |
| 43 | +- [#2155](https://github.com/docker/docker-agent/pull/2155) - fix: use dummy thought_signature for cross-model Gemini function calls |
| 44 | +- [#2156](https://github.com/docker/docker-agent/pull/2156) - Decline elicitations in run --exec --json mode |
| 45 | +- [#2157](https://github.com/docker/docker-agent/pull/2157) - Remove unused fileWriteTracker from creator package |
| 46 | +- [#2158](https://github.com/docker/docker-agent/pull/2158) - fix: use event timestamps for user messages in SessionFromEvents |
| 47 | + |
| 48 | + |
6 | 49 | ## [v1.32.5] - 2026-03-17 |
7 | 50 |
|
8 | 51 | This release improves agent reliability and performance with better tool loop detection, enhanced MCP handling, and various bug fixes. |
@@ -1394,3 +1437,5 @@ This release improves the terminal user interface with better error handling and |
1394 | 1437 | [v1.32.4]: https://github.com/docker/docker-agent/releases/tag/v1.32.4 |
1395 | 1438 |
|
1396 | 1439 | [v1.32.5]: https://github.com/docker/docker-agent/releases/tag/v1.32.5 |
| 1440 | + |
| 1441 | +[v1.33.0]: https://github.com/docker/docker-agent/releases/tag/v1.33.0 |
0 commit comments