Skip to content

CHANGELOG.md

Latest commit

 

History

History
334 lines (258 loc) · 9.55 KB

CHANGELOG.md

File metadata and controls

334 lines (258 loc) · 9.55 KB
Raw

🦅 Mohawk Inference Engine - Changelog

All notable changes to the project will be documented in this file.

[2.1.1] - 2026-07-11

🔧 Devcontainer and Setup Reliability

  • Updated .devcontainer/devcontainer.json:
    • workspace path compatibility for Codespaces
    • docker-outside-of-docker feature enablement
    • Python and Docker VS Code extension recommendations
    • explicit bash post-create invocation
  • Updated .devcontainer/Dockerfile with missing runtime and diagnostics packages.
  • Reworked .devcontainer/post_create.sh to be idempotent, root-safe, and configurable with:
    • MOHAWK_SKIP_LIBOQS_BUILD=1
    • MOHAWK_SKIP_PY_DEPS=1

🖥️ Full-Stack Desktop GUI Launch Improvements

  • Fixed startup ordering in mohawk_gui/main_window.py:
    • initialize status bar before tab refresh paths
    • start background health thread only after UI widget initialization
  • Added environment-based backend URL overrides for desktop GUI runtime.
  • Added reusable desktop GUI launcher flow in launch.py for both native and Docker orchestration.
  • Updated launch.sh dependency probe to include PyQt6.
  • Added mohawk-desktop-gui service to docker-compose.yml.
  • Added full-stack parity services to docker-compose.dev.yml and adjusted conflicting host port mapping.

📚 Documentation Updates

  • Updated README.md quick-start and documentation index.
  • Updated QUICKSTART.md with launcher/devcontainer paths and focused model/chat smoke checks.
  • Updated DOCKER_SETUP.md desktop GUI behavior notes and skip flag usage.
  • Added INSTALL.md and SETUP.md for clear install/runtime guidance.

✅ Validation

  • End-to-end functional test executed: python test_user_functions.py.
  • Result: 33/33 passed (100.0%) including model list/load and chat inference paths.

[2.1.0] - 2024-01-15

🎉 Major Release: Professional Dashboard with LM Studio Features

✨ New Features

Dashboard UI Complete

  • ✅ Implemented comprehensive tab-based interface with 7 main sections
  • ✅ Model Library Manager (LM Studio-style) with download/upload capabilities
  • ✅ Real-time Chat Interface with context management
  • ✅ Performance Metrics Dashboard with PyQtGraph charts
  • ✅ Session Manager with priority queue system
  • ✅ Worker Configuration with multi-device layer splitting
  • ✅ Security Center with PQC + mTLS + JWT authentication
  • ✅ Conversation History with usage analytics

📚 Model Library Features

Model Management

  • Browse available models with search and filters
  • Download models from Hugging Face or other sources
  • Upload local model files (.safetensors, .bin)
  • Quantization options: Q4_K_M, Q5_K_M, Q8_0, FP16
  • Device splitting configuration for multi-device inference
  • Model status tracking (Ready/Loading/Failed)

Model Details Panel

  • Load selected model with one click
  • Configure device split format: cpu_threads;gpu_ids
  • Real-time load progress indicator
  • Error recovery options on failure

💬 Chat Interface Features

Conversation Management

  • Multi-turn conversations with full context retention
  • Scrollable conversation history (500+ lines)
  • System prompt customization
  • Context size monitoring and management

Generation Controls

  • Temperature slider (0.0 - 2.0, default: 0.7)
  • Top-p sampling parameter (default: 0.9)
  • Max tokens setting (1 - 8192, default: 2048)
  • Send message with Enter key or Send button

Input Features

  • Placeholder text for guidance
  • Shift+Enter for new line
  • Auto-scroll to bottom on new messages
  • Clear history button for cleanup

📊 Performance Metrics Features

Real-time Charts (PyQtGraph)

  • Throughput chart: Requests per second
  • Latency percentiles: p50, p95, p99
  • Resource usage charts: CPU, Memory, GPU
  • Peak throughput tracking

Statistics Summary

  • Total requests processed
  • Average latency across sessions
  • Success rate percentage
  • Active sessions count
  • Peak throughput recorded

🔗 Session Manager Features

Session Table

  • Session ID with auto-generation
  • Model name display
  • Status indicators (Queued/Running/Completed)
  • Throughput metrics per session
  • Latency percentiles
  • Tokens per second
  • Start time timestamp
  • Cancel action per session

Queue Configuration

  • Max queue size spinner (1 - 1000, default: 50)
  • Priority queue buttons (High/Normal)
  • Visual status indicators by priority
  • Session lifecycle management

⚙️ Worker Configuration Features

Worker Management Table

  • Worker ID auto-generation
  • Host:Port configuration
  • Connection status (Connected/Disconnected)
  • Model name per worker
  • GPU thread count spinner
  • Load progress indicator
  • Connect/Disconnect actions
  • Restart failed workers

Multi-device Support

  • Device splitting configuration panel
  • Layer distribution across workers
  • GPU thread allocation per device
  • Load balancing monitoring

🔒 Security Center Features

JWT Authentication

  • Status indicator (Enabled/Disabled)
  • Algorithm display (RS256)
  • Token expiry time (24 hours default)
  • Refresh token button
  • Validation status

mTLS Configuration

  • Certificate validity check
  • Client certificate path
  • Key encryption status (Fernet)
  • Certificate renewal reminders

Post-Quantum Cryptography (PQC)

  • Hybrid KEM support toggle
  • X25519 + Kyber integration
  • liboqs status display
  • Enable/disable with confirmation
  • Quantum-resistant security layer

Security Event Log

  • Immutable audit trail
  • Timestamped events
  • Action type categorization
  • Resource tracking

📜 Conversation History Features

History Table

  • Timestamp for each conversation
  • Model used identifier
  • Tokens consumed count
  • Duration of conversation
  • Status (Completed/Failed)
  • Actions (view, delete, export)

Usage Statistics

  • Total tokens processed
  • Average latency across all sessions
  • Models usage count
  • Recent activity summary

🎨 UI/UX Improvements

Navigation

  • Tab-based interface for organized access
  • Navigation toolbar with icon buttons
  • Quick tab switching
  • Status bar with real-time updates

Visual Design

  • Professional color scheme (green success, orange warnings)
  • Emoji icons for quick identification
  • Bold status indicators
  • Consistent button styling
  • Rounded corners for modern look

Shortcuts

  • Ctrl+Shift+M: Show/Hide main window (tray icon)
  • Shift+Enter: New line in chat
  • Enter: Send message
  • F1: Show help

🔧 Technical Improvements

Architecture

  • Modular page design for maintainability
  • Event-driven architecture with signals/slots
  • Thread-safe metrics updates
  • Memory-efficient data structures

Dependencies

  • PyQt6 >= 6.5.0 (production-ready)
  • PyQtGraph >= 0.13.0 (fast charts)
  • cryptography >= 41.0.0 (security)
  • PyJWT >= 2.8.0 (token handling)
  • psutil >= 5.9.0 (system monitoring)

Code Quality

  • Comprehensive docstrings
  • Type hints for all functions
  • Error handling with graceful degradation
  • Input validation and sanitization

📚 Documentation Added

  • DASHBOARD_FEATURES.md - Complete feature documentation
  • QUICK_START.md - 3-minute setup guide
  • GUI_DASHBOARD_SUMMARY.md - Executive summary
  • Updated README.md with dashboard showcase
  • Inline code comments for all components

[2.0.0] - Previous Version

Core Features (Production Ready)

  • JWT Authentication with RSA signatures
  • mTLS Support for secure communication
  • Connection Pooling for high concurrency
  • Metrics Buffering with configurable windows
  • Graceful Error Handling with recovery strategies
  • Session Persistence and Checkpointing
  • Performance Monitoring (Memory, CPU, GPU)
  • Audit Logging for compliance

[1.0.0] - Initial Release

Prototype Features

  • Basic controller and worker architecture
  • Secure communication with X25519
  • Session management foundation
  • Telemetry hooks for monitoring

📋 Migration Guide from 2.0.0 to 2.1.0

Breaking Changes: None

The upgrade is seamless - all existing functionality is preserved while adding new dashboard features.

New Configuration Options

# Add to config.toml if needed
[dashboard]
enable_metrics_charts = true
metrics_update_interval_ms = 1000
max_history_items = 100

Migration Steps

  1. No code changes required - Dashboard is backward compatible
  2. Dependencies auto-resolve - PyQt6 and PyQtGraph will be installed if missing
  3. Existing models remain accessible - Model library loads all available models
  4. Worker connections preserved - Existing workers automatically connect

🎯 Roadmap for Future Versions

Version 2.2.0 (Planned)

  • Advanced chat history export (JSON, PDF)
  • Model comparison tools
  • Prompt template library
  • Batch inference support
  • Advanced analytics dashboard

Version 2.3.0 (Planned)

  • Plugin system for custom models
  • API endpoint management
  • Multi-user collaboration features
  • Real-time collaboration in chat

Version 2.4.0 (Planned)

  • Mobile companion app integration
  • Voice-to-text input
  • Advanced prompt engineering tools
  • Model distillation utilities

🙏 Acknowledgments

Inspiration

  • LM Studio - For user-friendly model interface design
  • Ollama - For model management concepts
  • vLLM - For performance optimization techniques

Contributors

Thank you to all contributors who helped build Mohawk Inference Engine!

📞 Support

For issues, questions, or contributions:

  • GitHub Issues: Open an issue
  • Documentation: See /docs directory
  • Quick Start: mohawk_gui/QUICK_START.md

Mohawk Inference Engine v2.1.0 - Professional Dashboard Released! 🦅