All notable changes to this project will be documented in this file.
- Hierarchical Pruning: Multi-stage file selection engine for repositories with 1,000+ files using Gemini 3 Flash.
- Explicit Metadata Enrichment: Integrated primary language and topic metadata from GitHub API for faster profile analysis.
- User-Selectable Reasoning Models: Introduced "Flash" and "Thinking" modes to allow users to choose between speed and deep reasoning.
- AI Pruning Efficiency: Switched selection logic to use Gemini 3 Flash Preview with "Minimal/Low Thinking" mode for rapid classification and lower latency.
- Profile Context Accuracy: Language detection now relies on authoritative GitHub metadata rather than README inference for core project tags.
- AI Core Cleanup: Removed unused Gemini 2.5 Flash Lite integration in favor of a unified Gemini 3 Flash architecture.
- Security Report UX: New DevTools Security tab with scan depth controls and AI toggle.
- File Scope Picker: Tree-based file selection to scope security scans precisely.
- Per-Finding Patch Copy: Generate and copy suggested patches for individual findings.
- Snippet Context: Findings now include surrounding code context for faster review.
- AI Quality Parsing: More resilient JSON parsing to prevent malformed AI responses from crashing analysis.
- Chat Export: Export chats to Markdown with Mermaid rendering.
- Copy Response: One-click copy for AI responses with enhanced clipboard support and icon updates.
- Chat UX: Reference text handling, selection support, and refined chat input/profile chat interfaces.
- File Preview: Refactored base64 decoding logic for readability.
- Markdown Support: User input contained between three backticks is now treated as markdown.
- PWA Support: Full Progressive Web App support is now available.
- GitHub Link: GitHub repository link is now implemented on the webpage.
- Blazing Fast Optimization: Instant profile loads & faster file selection with Flash-Lite.
- Native Web Search: Real-time answers for competitors, news, and jobs using Gemini's
googleSearch. - Smart URL Handling: Intelligent LinkedIn summarization & identity verification.
- Flowchart Accuracy: Increased accuracy in complex flowchart generation from 70% to 95+%.
- Error Handling: Mermaid syntax errors are now hidden from the UI.
- Context Awareness: Added context for better chat results.
- Code Blocks: Horizontal scrolling code blocks in chat UI.
- Enhanced Flowcharts: Integrated enhanced flowchart diagrams.
- Hidden Files: Added hidden files section to optimize the repo tree.
- DevTools: Implemented the devtools suite.
- Real-time Response: Added realtime Next.js server response.
- Caching: Introduced Vercel KV caching for GitHub API calls.
- Mermaid Fix Pipeline: Robust diagram fix pipeline using JSON-to-Mermaid conversion.
- Dev Tools Suite: Advanced Search (Regex/AST), Code Quality Analysis, and Automated Generators.
- Zero-Cost Security: Vulnerability scanning using pattern matching and Gemini AI.
- Enhanced Data: Precise language statistics via GraphQL and commit history.
- Auto-Persistence: Conversations are now automatically saved locally and restored instantly.
- Smart Caching: Instant load times for previously visited profiles and repositories.
- Mobile Experience: Optimized layout for mobile devices.
- Repo Chat: Deep dive into repositories with AI-powered chat and file analysis.
- File Preview: Instant syntax-highlighted previews for any file in the repository.