Release v4.0.0: Unified Multimodal Local AI Studio, 14-Tool MCP Server, and Verified Engine Control

[spelech]

🚀 Local LLM Server Manager v4.0.0 Release

Overview

This release elevates Local LLM Server Manager to v4.0.0, delivering a unified multimodal local AI orchestrator with end-to-end verified engine control across all modalities (Text, Image, Audio/Voice, 3D Mesh, and Video), an expanded 14-tool Model Context Protocol (MCP) server, architectural refactoring to enforce single-responsibility classes, and complete test suite integrity.

---

🌟 Key Highlights & Changes

1. Architectural Refactoring & Single Responsibility

• Separated Multi-Class Files:
  • EngineStudioViewModel.cs refactored from 900+ lines down to ~300 lines by extracting MainViewModel.Studio.cs (handling prompt docks, style presets, and Forge dispatch).
  • MainViewModel.cs trimmed from 1,600+ lines to ~1,200 lines by extracting MainViewModel.Generation.cs (multi-modal execution commands for Ollama, 3D mesh, Kokoro audio, and ComfyUI video).
  • Extracted UI converters (StickerStudioConverters.cs, AttachmentToImageConverter.cs) into Converters/.
  • Extracted DTO records and models into dedicated files: AudioStudioModels.cs, DocumentationModels.cs, HubModelItems.cs, WindowSnapModels.cs, and WorkflowRequestModels.cs.
• Enforces strict one-class-per-file responsibility across all projects.

2. Verified Engine Control & Automated Testing Suite (Zero Test Theatre)

• Added genuine integration and execution tests verifying actual engine API contracts and payloads:
  • EngineCommandExecutionTests.cs: Real payload verification for Ollama (/api/generate), Forge (/sdapi/v1/txt2img), Kokoro (/v1/audio/speech), and ComfyUI (/prompt, /api/3d/generate).
  • HuggingFaceProviderTests.cs & CivitaiProviderTests.cs: Mocked provider API tests and headless UI interaction tests for model card discovery and inspection.
  • ComfyWorkflowGenerationTests.cs: Graph translation, node parameter injection (CFG, steps, denoise, seed), and cancellation checks.
  • ForgeProviderTests.cs: State machine polling (0% -> sampling steps -> 100%), checkpoint discovery/switching, aspect ratio resolution, and headless UI progress updates.
  • KokoroVoiceProviderTests.cs: Voice catalog parsing, audio streaming, automatic fallback injection (af_heart), and live playback states.
  • Live External Engine Probing: Health and integration probes across live ports (11434, 7860, 8188, 8880).
• Test Results: 924 passed, 18 skipped, 0 failed (942 total).

3. Expanded 14-Tool Model Context Protocol (MCP) Server

• Extended LocalLlmMcpTools.cs and /mcp HTTP/SSE endpoint to support 14 native tools:
  • get_gpu_vram, check_health, list_models, pull_model, unload_vram, start_engine, stop_engine, detect_tools
  • Creative & Multimodal: generate_video, synthesize_speech, generate_audio
  • Hub Discovery & Workflows: search_huggingface, search_civitai, run_studio_workflow

4. Forge Engine Readiness UX

• Added proactive offline warning badge in the Image generation studio when Forge port 7860 is down.
• Integrated a 1-click Start Forge Engine button directly within the warning banner for seamless recovery.

5. Documentation & Metadata Sweep

• Solution-wide version bump to v4.0.0 across desktop .csproj, WASM web project, installer scripts, Inno Setup config, UI views, tray tooltip, and documentation.
• Updated README.md, docs/index.md, and docs/ai-and-mcp/mcp-tools.md to reflect v4.0.0 architecture and the 14-tool MCP catalog.
• Fully verified with npm run lint, npx tsc --noEmit, and dotnet test.