v0.77.0
AiderSummary
This release introduces powerful new configuration commands, /think-tokens and /reasoning-effort, allowing fine-grained control over model behavior and budget. It also brings massive expansion to supported programming languages via tree-sitter integration and fixes several critical bugs related to model configuration and file handling.
New Features
- Added /think-tokens command to set thinking token budget with support for human-readable formats (8k, 10.5k, 0.5M).
- Added /reasoning-effort command to control model reasoning level.
- Enabled support for Cohere model cohere_chat/command-a-03-2025 and Gemini model gemini/gemma-3-27b-it.
- Added --auto-accept-architect flag to automatically accept changes from architect coder format without confirmation.
Bug Fixes
- Fixed a bug where the default model was incorrectly set by deprecated --shortcut switches even when explicitly specified on the command line.
- Fixed handling of JSONDecodeError when loading the model cache file.
- Fixed handling of GitCommandError when retrieving git user configuration.
Improvements
- Expanded programming language support significantly by adopting tree-sitter-language-pack, adding 130 new languages with linter support and 20 new languages with repo-map support.
- The /think-tokens and /reasoning-effort commands now display current settings when called without arguments.
- Display of thinking token budget and reasoning effort is now available in model information.
- Changed the --thinking-tokens argument to accept string values using human-readable formats.
- Improved AutoCompleter to require 3 characters for autocompletion to reduce noise.
- The bare /drop command now preserves original read-only files provided via args.read.
- Added ignore_permission_denied option to the file watcher to prevent errors when accessing restricted files.