Tags: athem135-source/PCBot
Tags
Fix truncation: restore strict 45-70 word limit and prevent currency… …/number cutoffs - Restored WORD_LIMIT=70 for Groq responses - Appends numeric tokens when truncation would cut currency mentions (e.g., 'Rs. 10,000') - Ensures sentence termination and clean citations Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
v3.4.0 Major Release: GitHub Pages, Mode Selector, Venv Automation, E… …nhanced UI User Interface Enhancements: - Mode selector landing page (User/Admin modes) - Server-side password authentication - Smooth animations (slideUp, fadeIn, bounce effects) - Backend health check with retry overlay Deployment Options: - GitHub Pages auto-deployment via Actions - Netlify deployment configuration - Enhanced Cloudflare tunnel with retry warnings - Multi-platform static hosting ready Setup & Installation: - Virtual environment auto-creation (.venv) - One-command setup via setup.bat - Embedding model auto-download on warmup - Auto-start Qdrant and Ollama services - Root-level launcher visibility Professional Structure: - public/ folder for static files - scripts/setup/ for installation - deployment/ for hosting guides - Clean separation of concerns Security Improvements: - Server-side admin authentication - Session-based access control - No client-side password exposure - Enhanced CORS for multi-domain Documentation: - QUICKSTART.md for new users - GITHUB_PAGES.md deployment guide - NETLIFY.md deployment guide - Comprehensive README updates
refactor: Organize test files into proper subfolders
- Created tests/results/ for JSON test result files
- Created tests/reports/ for markdown report files
- Updated comprehensive_300_test.py to save results to tests/results/
- Updated generate_report.py to read from tests/results/ and save to tests/reports/
- Updated run_calibration_test.bat with new paths
- Added proper type annotations to fix Pylance warnings
- Added os import for path operations
File structure:
tests/
results/ <- JSON test results
reports/ <- Markdown reports
*.py <- Test scripts
feat: Add 300-question calibration test suite with detailed documenta… …tion - Created comprehensive_300_test.py with 9 test categories (300 questions) - Added generate_report.py for detailed markdown test reports - Added run_calibration_test.bat for one-click test execution - Enhanced multi_classifier.py with 15+ new red-line detection patterns - Updated README.md with full test results and collapsible Q&A sections Test Results: 278/300 passed (92.7% accuracy) - In-Scope Detailed: 100% (50/50) - In-Scope Short: 100% (50/50) - Numeric/Financial: 100% (40/40) - Trap Questions: 100% (30/30) - Trick Questions: 100% (30/30) - Off-Scope: 100% (40/40) - Greetings: 100% (15/15) - Manual Sections: 100% (20/20) - Red-Line Safety: 12% (3/25) - classifier patterns enhanced post-test
v3.3.4: Full chat image export, enhanced download options - Fixed image download to capture full chat history (not just visible area) - Temporarily expands scrollable container before capture - Added jsconfig.json to fix html2canvas TypeScript warning - Updated all version numbers to 3.3.4 - Enhanced download options: Text, HTML (styled), Image (PNG)
feat: Enhanced download options - HTML styled chat + Image export Download Options: - Text (.txt): Plain text export - HTML (.html): Styled chat interface that looks like the widget - Image (.png): Screenshot of chat using html2canvas HTML Export Features: - Gradient header matching widget design - User/bot message bubbles with proper styling - Timestamps on each message - Print-friendly CSS - Session metadata in footer Image Export: - High quality (2x scale) - Captures entire chat body - Uses html2canvas library Updated components: - storage.js: Added exportChatAsHTML, downloadAsImage - ChatWidget.jsx: New download handlers - SettingsMenu.jsx: Updated menu with 3 download options - Added html2canvas dependency
PreviousNext