v0.9.73 @magnitudedev/cli@0.2.2
vadimdemedes/inkSep 30, 2026 → Oct 1, 20260 days between releases
What's New
Fixes critical stability issues in inference engine and WebSocket communication
New Features
- Improved error handling for tool calls across multiple AI models
- Optimized memory bandwidth estimation for model assessment
- Improved caching of large system prompts
Improvements
- Optimized memory bandwidth estimation for model assessment
- Improved caching of large system prompts
- Improved error handling for tool calls across multiple AI models
- Fixed token usage counting for Anthropic models
Bug Fixes
- Fixed app getting stuck on 'Assessing models' by estimating speed via memory bandwidth
- Fixed Codex hanging after tool calls by properly managing request output state
- Fixed tool parameter parsing errors for Gemma 4
- Fixed repeated image encoding and tool ID generation
- Fixed large system prompts being re-read in full when only the last message changes
Upgrade Notes
No configuration changes are required; this is a pure stability patch. Ensure Anthropic API keys have max_tokens set for token counting to work correctly.
The Two Releases
v0.9.73
Sep 30, 2026· 22 days ago
A patch release for graphify adding support for extracting enums and free functions in Rust, Zig, C++, and Scala 3, alongside various parsing and community labeling fixes.
Full release notes@magnitudedev/cli@0.2.2
Oct 1, 2026· 21 days ago
This patch release addresses critical stability issues in the inference engine, fixing hanging behaviors during model assessment, WebSocket communication, and complex tool parsing for various AI models.
Full release notes