Changelog - jmpijll/discomfy GitHub Wiki

Changelog

Complete version history for DisComfy.

[2.0.0] - November 2, 2025

🎉 Major Release: Complete Architectural Overhaul

This is the biggest release in DisComfy history, featuring a complete transformation from monolithic to modular architecture while maintaining 100% backward compatibility.

🏗️ Architecture

  • Refactored: Complete transformation from monolithic to modular architecture
  • Structure: Organized into bot/, core/, config/, utils/ directories
  • Reduction: 77% reduction in max file size (3,508 → 705 lines)
  • Modularity: 50+ well-organized files with clear separation of concerns
  • Entry Point: New main.py with clean architecture

🎯 Code Quality

  • Testing: Implemented comprehensive test suite (85/86 tests passing, 99% pass rate)
  • Best Practices: Following Context7 patterns for discord.py and aiohttp
  • Type Safety: Full Pydantic V2 migration with type hints throughout
  • Design Patterns: Strategy, ABC, Factory, Repository patterns implemented
  • Validation: Pydantic-powered validation with clear error messages

✨ Features

  • Commands: All 6 commands refactored with new architecture
    • /generate - Interactive generation with model selection
    • /editflux - High-quality Flux Kontext editing
    • /editqwen - Ultra-fast Qwen 2.5 VL editing
    • /status - Bot and ComfyUI status
    • /help - Command help
    • /loras - List available LoRAs
  • Generators: Refactored ImageGenerator and VideoGenerator to v2.0 architecture
  • Progress: Simplified progress tracking with better accuracy
  • UI Components: Modular Discord UI (buttons, modals, views)

🐛 Bug Fixes

  • WebSocket: Continued improvements from v1.4.0
  • Progress: Fixed 100% completion display issues
  • Pydantic: Removed all V1 deprecation warnings
  • Discord.py: Fixed SelectOption import for 2.x compatibility
  • Video: Fixed Request/Response pattern issues
  • Validation: Enhanced workflow validation

📚 Documentation

  • Created: 24+ comprehensive documentation files
  • Guides:
    • Migration Guide (v1.4.0 → v2.0.0)
    • API Reference (complete API docs)
    • Usage Examples (practical examples)
    • Testing Guide (comprehensive testing docs)
    • Installation Guide (Standard, Docker, Unraid)
    • Configuration Guide (all options explained)
    • Troubleshooting (common issues and solutions)
  • Organization: Archived historical docs, clean root structure
  • Release Notes: Complete v2.0.0 release documentation

🐳 Docker & Deployment

  • Dockerfile: Updated for v2.0 structure, uses main.py
  • GitHub Actions: Configured for both GHCR and Docker Hub
  • Containers: Auto-publish to:
    • ghcr.io/jmpijll/discomfy
    • jamiehakker/discomfy (Docker Hub)
  • Optimization: Better layer caching, smaller image size
  • Tags: Latest, version-specific, and commit SHA tags

🔄 Backward Compatibility

  • 100% Compatible: No breaking changes from v1.4.0
  • Migration: Old entry points still work (bot.py)
  • Configuration: No config changes required
  • Workflows: All workflows compatible
  • Commands: All commands function identically

⚡ Performance

  • Startup: ~1 second (well under 3s target)
  • Tests: 2.58s for 86 comprehensive tests
  • Code: 60-75% reduction in key modules
  • Memory: Optimized resource usage

🛠️ Developer Experience

  • Imports: Organized imports from new modules
  • Testing: Easy to test individual components
  • Extensibility: Clean base classes for new features
  • Type Hints: Full type coverage
  • Linting: Passes all quality checks

📊 Statistics

  • Code Size Reduction: 77% (3,508 → 705 lines max)
  • Test Pass Rate: 99% (85/86 tests)
  • Files Created: 50+ organized modules
  • Documentation: 24+ comprehensive guides
  • Commits: 100+ commits in v2.0 development

[1.4.0] - October 31, 2025

🐛 Critical Bug Fix: Concurrent Generation Hanging Issue Resolved

Fixed

  • Concurrent Generation Hanging Bug: Fixed critical issue where second concurrent generation would complete in ComfyUI but never display in Discord
    • Moved WebSocket initialization from per-generation to bot startup
    • Removed context manager pattern that caused race conditions
    • Added explicit initialize() and shutdown() methods
    • WebSocket now opens once and stays alive for entire bot lifetime
    • Both concurrent generations now complete successfully without hanging
    • Fixes #2

Improved

  • Performance: 5-10x faster generation queue times (no session recreation)
  • Reliability: Automatic WebSocket reconnection (up to 999 retries)
  • Code Quality: Removed 47 lines of complex code, simpler architecture
  • Resource Management: Single persistent HTTP session for bot lifetime

Technical Details

  • image_gen.py: Removed context managers, added lifecycle methods (~150 lines modified)
  • bot.py: Updated initialization, removed 14 context manager usages (~20 lines modified)
  • video_gen.py: Added lifecycle methods with resource sharing (~12 lines modified)

Testing

  • ✅ Verified on production server with concurrent generations
  • ✅ All tests passing, no regression
  • ✅ WebSocket reconnection tested
  • ✅ Startup and shutdown lifecycle tested

[1.3.1] - October 5, 2025

🐛 Critical Bug Fixes for Custom Workflows

Fixed

  • LoRA Selector Error: Fixed 'CompleteSetupView' object has no attribute 'insert_item' error

    • Changed from non-existent insert_item() to correct children.insert() method
    • LoRA selector now initializes properly without errors
    • Credit to @AyoKeito for identifying the issue

  • Custom Workflow Validation: Fixed "Cannot execute because a node is missing the class_type property" error

    • Added comprehensive _validate_workflow() method to check workflow structure
    • Validates all nodes have required class_type property
    • Checks for proper workflow dictionary structure
    • Warns about missing inputs properties

Added

  • Intelligent Error Messages: New validation provides detailed, actionable error messages
    • Lists specific nodes with issues (up to 5)
    • Explains what's wrong with each node
    • Provides common solutions and export format guidance
    • Tells users to use "Save (API Format)" in ComfyUI
    • Includes example of correct node structure

Technical Details

  • bot.py (Line 823): Fixed LoRA selector initialization method
  • image_gen.py (Lines 438-511): Added _validate_workflow() method
  • image_gen.py (Line 429): Updated _load_workflow() to call validation
  • Validation runs before sending workflow to ComfyUI, catching errors early

Impact

  • ✅ Custom workflows now provide clear guidance when structure is invalid
  • ✅ Users can debug workflow issues without trial-and-error
  • ✅ LoRA selector works reliably for all model types
  • ✅ Better developer experience for custom workflow creation

[1.3.0] - October 4, 2025

✨ New Feature: Multi-Image Qwen Editing

Added

  • Multi-Image Qwen Edit: /editqwen command now supports 1-3 input images
  • Workflow Selection: Automatically selects correct workflow based on image count:
    • 1 image → qwen_image_edit.json
    • 2 images → qwen_image_edit_2.json
    • 3 images → qwen_image_edit_3.json
  • Optional Image Parameters: Added image2 and image3 optional parameters to /editqwen command
  • Smart Validation: Ensures image3 cannot be provided without image2

Changed

  • Image Generator: Enhanced generate_edit() method with additional_images parameter
  • Workflow Parameters: Updated _update_qwen_edit_workflow_parameters() to handle multiple LoadImage nodes
  • Progress Display: Shows total image count in progress messages
  • Initial Response: Displays all input images with their file sizes

Technical Details

  • Multi-image data passed as List[bytes] via additional_images parameter
  • Images uploaded with unique timestamps to ComfyUI
  • LoadImage nodes automatically assigned to correct images by node ID order
  • All Qwen workflows (1, 2, and 3 images) fully supported

[1.2.11] - October 4, 2025

🔧 Critical Fix: True Persistent WebSocket Implementation

Fixed

  • Concurrent Generation Completion: Fixed bug where second generation never completed until third started
  • Progress Tracking: Fixed issue where progress stayed on "Checking status..." then jumped to 100%
  • WebSocket Architecture: Implemented single persistent WebSocket that monitors ALL generations simultaneously
  • Session Management: WebSocket now stays alive across multiple async with context manager cycles

Technical Details

  • Previous fix (v1.2.10) used shared client_id but still created separate WebSocket per generation
  • Multiple WebSockets competed for messages, causing progress tracking failures
  • New architecture: ONE persistent WebSocket for entire bot session
  • Each generation registers/unregisters with _active_generations dictionary
  • Persistent monitor routes messages to correct generation by prompt_id
  • WebSocket starts on first generation, persists across all subsequent generations

Impact

  • ✅ All concurrent generations now complete correctly without manual intervention
  • ✅ Real-time progress tracking works for all generations
  • ✅ No more "stuck at Checking status" issues
  • ✅ True concurrent queue handling as originally intended

[1.2.10] - October 3, 2025

🐛 Critical Queue Handling Fix

Fixed

  • Queued Generation Progress: Fixed critical bug where generations started while another is running would never show progress percentages
  • WebSocket Client Filtering: Removed ?clientId parameter from WebSocket connections
  • Completion Detection: Fixed issue where completed generations weren't detected until new generation started
  • Progress Message Filtering: Implemented proper prompt-based filtering
  • Concurrent Request Handling: Multiple users can now safely queue generations simultaneously

Technical Details

  • Removed ?clientId={client_id} from WebSocket URL that was filtering messages
  • WebSocket now receives ALL messages from ComfyUI
  • Added currently_executing_prompt tracking
  • Progress messages filtered by matching against currently executing prompt_id

[1.2.9] - October 3, 2025

🐛 Concurrent Generation Progress Tracking Fix

Note: This version exists in remote repository with different implementation approach.

[1.2.8] - October 3, 2025

🚀 Dual Image Editing Models

Added

  • Qwen 2.5 VL Ultra-Fast Editing: New /editqwen command (30-60 seconds)
  • Dual Edit Buttons: Both ✏️ Flux Edit and ⚡ Qwen Edit buttons on generated images
  • Smart Parameter Validation: Context-aware step validation (4-20 for Qwen, 10-50 for Flux)
  • Automatic CFG Adjustment: Optimized CFG values per workflow
  • New Workflow Integration: Added qwen_image_edit.json workflow support

Changed

  • Command Renamed: /edit → /editflux (Flux Kontext high-quality editing)
  • Edit Buttons Split: Single edit button now split into Flux and Qwen variants
  • Modal Titles: Context-aware modal titles indicating model being used
  • Progress Messages: Model-specific progress indicators

Earlier Versions

For complete version history, see the repository's commit history.

Version Numbering

DisComfy follows Semantic Versioning:

  • MAJOR version (2.0.0) - Incompatible API changes (but v2.0.0 is backward compatible!)
  • MINOR version (1.3.0) - New functionality in backward-compatible manner
  • PATCH version (1.3.1) - Backward-compatible bug fixes

Upgrade Guide

See Migration Guide for detailed upgrade instructions.

Quick Upgrade:

git pull origin main
pip install -r requirements.txt --upgrade
python main.py

For the latest changes, see GitHub Releases.