Tags: Flux159/mcp-game-asset-gen
Tags
Update README to reflect Gemini 3 Pro model support - Update gemini_generate_image tool description to mention both 2.5 Flash and 3 Pro models - Clarify that 'gemini' model option now defaults to gemini-3-pro-image-preview - Update all references to Gemini model defaults throughout documentation
Add MCP transparency integration test - Test verifies generate_texture tool handles transparentBackground parameter - Confirms transparency conversion works with light grey to white (#efefef to #ffffff) - Validates auto background detection functionality - Demonstrates end-to-end MCP transparency workflow
Implement actual transparency conversion functionality - Replace placeholder convertToTransparentBackground with real pixel conversion - Add canvas-based pixel manipulation for transparency - Support white, black, and auto background detection with tolerance - Add comprehensive tests for transparency conversion - Install canvas package for image processing
Remove synchronous 3D generation to prevent MCP timeouts and improve … …async documentation - Remove image_to_3d synchronous tool that causes 60-second MCP timeouts - Keep only image_to_3d_async as the primary 3D generation method - Improve async tool documentation with detailed status file format - Add comprehensive monitoring usage examples and progress stages - Update server tests to reflect 10 tools total (removed sync tool) - Ensures users 'fall into the pit of success' by only exposing reliable async method This prevents frustrating timeout errors for all 3D generation operations since they typically take longer than the MCP protocol timeout limit.
Fix Hunyuan3D Multi API field requirements and add validation - Update hunyuan3DGenerateMulti to use front_image_url, back_image_url, left_image_url fields - Update hunyuan3DGenerateMultiTurbo to use specific view fields instead of image_urls array - Add validation to ensure at least 3 images are provided for multi-view APIs - Fix corrupted hunyuan3DGenerateSingleTurbo function that was mixed with multi-image code - Improve error messages to be more descriptive about image requirements This fixes the API error: 'field required' for front_image_url, back_image_url, left_image_url
Implement enum validation for 3D model generation to prevent API errors - Add Model3DModel, Model3DVariant, and Model3DFormat enums - Create AVAILABLE_VARIANTS mapping to prevent invalid model/variant combinations - Update tool schema to use Object.values() of enums for dynamic validation - Fix FAL.ai API endpoints for Hunyuan3D variants - Update all tests to use proper enum values instead of string literals - Re-export enums from model3dHelpers for use in index.ts This ensures users 'fall into the pit of success' by only exposing valid options in the tool schema, preventing API endpoint errors from invalid combinations.
Fix CD workflow to handle detached HEAD state - Add explicit checkout of main branch before updating versions - Pull latest changes from main to avoid conflicts - Fix 'src refspec main does not match any' error when triggered by tag push - Ensure version updates are committed to the correct branch
Add CI/CD workflows for automated testing and npm publishing CI (.github/workflows/ci.yml): - Runs on pull requests and pushes to main - Uses Node.js 20 with npm caching - Runs full test suite with JUnit reporting - Includes type checking and build verification - Tests npm packaging with npm pack --dry-run - Uses test-reporter action for GitHub check results CD (.github/workflows/cd.yml): - Runs only on version tags (v*) - Runs full test suite and type checking before publishing - Builds production version automatically - Publishes to npm using NPM_TOKEN secret - Creates GitHub releases with auto-generated notes - Sets up Node.js with npm registry authentication Ready for automated releases!