Repository navigation
1.99.30058
·
97 commits
to main
since this release
A-Coder 1.99.30058
🚀 New Features
- Expanded Skill System: Introduces execution, evaluation, and marketplace capabilities for skills—enabling users to discover, run, and track mastery of AI-assisted learning modules directly in the sidebar.
- Multi-Workspace Management: Unified workspace selection, cross-workspace thread search, and aggregated statistics in the Agent Manager—ideal for developers managing multiple projects.
- Browser & Webview Tools: New tools for AI-driven browser interaction and webview content manipulation (e.g., form submission, DOM inspection), expanding A-Coder’s automation scope.
- Learning & Progress Tracking: Spaced-repetition quizzes, skill mastery dashboards, and inline exercise validation—designed to help users learn and retain coding concepts.
- What’s New Modal: Interactive onboarding and feature highlights for new releases (WIP).
- Enhanced Image Generation: Moved image generation to the main process via IPC to support self-signed SSL certificates; default model updated to
x/flux2-klein:4b. - Improved Edit Tool: Simplified
edit_filetool using direct string replacement (no moreORIGINAL/UPDATEDblocks), with robustness improvements like concurrent edit protection, content validation, and line-ending normalization.
🐛 Bug Fixes
- Chat Rendering Stability: Fixed out-of-order message display by enforcing monotonic raw text and consistent partitioning of
thinktags in streaming responses. - Initial Scroll Position: Corrected chat sidebar to show newest messages first in virtualized lists.
- Markdown Streaming: Prevented malformed markdown display by rendering plain text during streaming, handling incomplete code blocks, and adding fallback rendering for malformed structures.
- Image Rendering: Fixed broken images by using full file URIs instead of relative paths.
- Emoji Rendering: Resolved display inconsistencies by replacing legacy Unicode with explicit
\u{...}escapes and improving rendering logic across learning and chat components.
🔧 Improvements
-
Performance & Memory:
- Removed virtualized chat rendering in favor of scroll-based virtualization for better large-thread performance.
- Added memory optimizations (e.g., delayed OAuth service instantiation, component memoization, and leak prevention).
- Increased UI update frequency for smoother streaming experience.
-
AI Streaming & Responsiveness:
- Enabled real-time streaming of LLM responses and reasoning deltas.
- Improved markdown rendering stability with content-based keys and refined handling of incomplete structures.
-
Text Matching & Editing:
- Replaced n-gram similarity with Levenshtein distance for more accurate fuzzy matching.
- Added trailing whitespace/tab normalization to improve edit detection reliability.
-
Accessibility & UI:
- Added screen reader support and keyboard navigation enhancements in Agent Manager and sidebar components.
- Redesigned UI with premium styling, VS Code design tokens, and theme-aware icons (e.g.,
VoidIconwith theme inversion). - Refined emoji, quiz results, walkthroughs, and implementation plan previews.
-
Tooling & Configuration:
- Updated OLLAMA model to
glm-5:cloudin release notes generator. - Re-enabled A-Coder provider with API key authentication (OAuth temporarily disabled for stability).
- Updated dependencies and refactored markdown rendering components for maintainability.
- Updated OLLAMA model to
📝 Other Changes
-
Branding & Documentation:
- Overhauled README.md, tool architecture docs, and codebase guides to reflect A-Coder branding.
- Removed obsolete icons, assets, and outdated documentation.
- Revised A-Coder Provider API spec and added fine-tuning documentation for tool orchestrators.
-
Telemetry & Compliance:
- Implemented comprehensive telemetry for sessions, files, and UI interactions.
- Updated copyright notices and improved terminal command correlation with shell integration.
Release generated by Hamish