AI Development Tools and Integration Ecosystem

Tools for AI-assisted development, image generation workflows, training monitoring, and intelligent task management. From ComfyUI optimization to AI-powered file curation.

DazzleML brings together tools for working with AI systems - whether you're generating images, training models, organizing datasets, or building AI-integrated workflows. These tools emerged from personal AI development work and help to solve problems encountered in various environments.

Read our philosophy → | See who should use DazzleML →

Most Popular Dazzle Project Overall

One-click installation of Triton and SageAttention optimizations for ComfyUI. Get 20-40% speed improvements without manual compilation.

python comfyui_triton_sageattention.py --install  # Everything installs automatically

Full documentation → | Status: Active

Image deduplication and quality-based selection tool. Scans directories for similar images using CLIP embeddings, evaluates them with objective quality metrics (dimensions, resolution, filesize, format), and organizes the highest-quality version of each group.

python find_best_images.py -i ./photos -i ./backup -o ./best -r --copy-best

Full documentation → | Status: Active

Monitor and visualize AI model training in real-time. Track loss, learning rate, sample outputs, and system resources.

python train.py --monitor localhost:8080
# View at http://localhost:8080

Full documentation → | Status: Active

Tools for integrating AI assistants into development workflows using the Model Context Protocol. Includes TodoAI integration.

Full documentation → | Status: Active

# 1. Optimize ComfyUI for faster generation
python comfyui_triton_sageattention.py --install

# 2. Generate images (now 20-40% faster)
# Use ComfyUI normally

# 3. Deduplicate and select best-quality versions
python find_best_images.py -i outputs/ -o best/ -r --copy-best --collect-results

# 4. Monitor training runs
python train.py --monitor localhost:8080

• Who Should Use DazzleML - Artists, developers, researchers, studios
• Philosophy - Design principles and approach
• Roadmap - Current status and planned features
• Sustainability - Supporting development

Project Documentation:

• ComfyUI Triton Installer
• find-best-images
• AI Training Monitor
• MCP Integrations

DazzleML works seamlessly with other Dazzle organizations:

DazzleLib - Use dazzle-filekit for dataset verification and file operations

DazzleTools - Use preserve for training data backups, pattern-break for dataset organization

DazzleNodes - ComfyUI custom nodes optimized with Triton installer

Contributions welcome! Each tool has its own repository with contribution guidelines.

General Process:

1. Fork the tool repository
2. Create a feature branch
3. Test with real AI workloads
4. Ensure cross-platform compatibility
5. Submit a pull request

Read contributing guidelines →

Learn about supporting DazzleML →

If DazzleML saves you time, consider sponsoring on GitHub or buying us a coffee

Current sponsorship: $0/month | Goal: $1,000/month

DazzleML is one of five organizations in the Dazzle ecosystem:

• DazzleProj - Ecosystem coordination
• DazzleLib - Foundation libraries
• DazzleTools - Command-line tools
• DazzleNodes - ComfyUI custom nodes
• DazzleML - AI development tools ← You are here

• TodoAI - Task-based OS with AI integration
• ComfyUI - Node-based UI for Stable Diffusion
• Model Context Protocol - AI assistant integration standard

DazzleML projects use GPL 3.0 to encourage open AI tooling and ensure improvements benefit everyone.

See individual project repositories for specific licensing information.

• GitHub: @djdarcy
• Issues: Use repository-specific issue trackers
• Discussions: GitHub Discussions
• Sponsorship: GitHub Sponsors
• Website: DazzleProj.com (coming soon)

Built by someone who trains models and generates images every day 🤖

AI tools for real work, not just demos.