v0.4.1
Session recovery, anonymous diagnostics, and custom models
Continue sessions after long responses are truncated, opt out of privacy-safe anonymous error reports, and validate local StartLux Qwen GGUF models before using them.
What's new
- Added StartLux custom models: place one Qwen GGUF and an optional mmproj vision file in the designated folder, then pass real text and image tests before making it the default model.
- Fixed sessions failing to continue after a long response reaches the output limit; generated content is preserved and sending ‘Continue’ resumes the session.
- Added anonymous error reports for Agent, inference-engine, and client crashes. Reports contain only allowlisted structured diagnostics—never conversations, prompts, file contents, paths, or raw logs—and can be disabled in Settings.
- Local models now use a bounded reasoning budget by default and support turning reasoning off, reducing the chance of exhausting the output allowance.