Changelog
All major updates and releases.
April 14, 2025
Beta Patch #4
v1.2.1
- Added checkpoints that let you jump between LLM edits.
- Agent mode can now fix lint errors after editing files.
- Upgraded tool-calling implementation; now, any model can run in Agent mode - R1, Gemma3, Quasar Alpha, etc.
- Dynamic context squashing when the context window is not sufficiently large.
- Added SSH support, WSL support, and Linux support.
- Void now comes with auto-updates, so we'll be pushing smaller and more frequent changes.
- GPT 4.1, Gemini 2.5 Pro, OpenHands LM, DeepSeek V3, Phi4, and Quasar Alpha support.
April 7, 2025
Beta Patch #3
v1.0.3
- Experimental version of Void pushed to our Discord members.
- Initial version of v1.1.0 above.
March 22, 2025
Beta Patch #2
v1.0.2
- Added Agent mode, Gather mode, and Chat mode.
- Agent mode can control your terminal, read/write files, and search your codebase.
- Chat and Gather mode now make suggestions to edit specific files.
- Void auto-detects when a model supports tools (Agent/Gather), FIM (Autocomplete), and thinking (Reasoning).
- New slider to set a model's reasoning level preference.
- Autocomplete has been re-enabled, and you can use it with any FIM model.
- Added a Fast Apply option, enabled by default.
- Chat mode now creates links to symbols in your code.
- Rebase from VSCode 1.99.0.
- Claude 3.7, Deepseek V3, Gemini 2.0, and QwQ support.
January 23, 2025
Beta Patch #1
v1.0.1
- New default theme.
- Add support for DeepSeek.
- Fixed system prompt errors when using o1 and Gemini.
- Improved prompts for fast-apply and quick edit.
- Performance improvements for chat.
- Temporarily disable autocomplete (needs better model selection guardrails).
- Minor updates to diff streaming algorithm and FIM output parsing algorithm.
January 19, 2025
Beta Release
v1.0.0
- Added quick edits (Ctrl+K)! This includes FIM-prompting, output parsing, and history management.
- Added autocomplete!
- Void now lives natively in the VSCode codebase, no more extension API.
- Added new UI to smoothly show the LLM's streamed changes in VSCode.
- New settings page with one-click switch, model selection, and even more providers.
- Added auto-detection of Ollama models by default.
- Fixed CORS and CSP errors for many local models by originating LLM calls from 'node/' with IPC.
- Native UI for Accept/Reject buttons, streaming, and interruptions.
- File suggestions in the chat based on history.
- Switched from the MIT License to the Apache 2.0 License for a little more protection on our OSS initiative.
October 1, 2024
Early Launch
- Initialized Void's website and GitHub repo.
- A lot of early setup work not recorded in this changelog.
- Basic features like LLM streaming in the editor, custom history, custom editor UI.
- Users could build Void for an early version of Ctrl+L and fast-apply.