MaIN.NET.0.10.16

0.10.14 release

• Local image generation now runs in-process via StableDiffusion.NET (GGUF diffusion models on CPU/CUDA), replacing the external Python/Flask image-gen service.
• Adds local diffusion model catalog (StableDiffusion1_5, Flux1Shnell, QwenImage) with multi-asset downloads (VAE/CLIP/T5/etc) via AIHub.Model().EnsureDownloadedAsync().
• InferPage Settings now lists local diffusion models for the Self backend, tagged "Image Gen", and resolves custom model paths correctly for image generation.
• Removes the ImageGen Python service, Docker image, and MaIN__ImageGenUrl setting.

MaIN.NET.0.10.14

0.10.14 release

• Local image generation now runs in-process via StableDiffusion.NET (GGUF diffusion models on CPU/CUDA), replacing the external Python/Flask image-gen service.
• Adds local diffusion model catalog (StableDiffusion1_5, Flux1Shnell, QwenImage) with multi-asset downloads (VAE/CLIP/T5/etc) via AIHub.Model().EnsureDownloadedAsync().
• InferPage Settings now lists local diffusion models for the Self backend, tagged "Image Gen", and resolves custom model paths correctly for image generation.
• Removes the ImageGen Python service, Docker image, and MaIN__ImageGenUrl setting.

MaIN.NET.0.10.13

0.10.13 release

• Adds WithMaxIterations() to ToolsConfigurationBuilder and McpContext to override the default tool-call iteration limit.
• Fix: MCP loop now sends a final synthesis request instead of returning an error string when the iteration cap is reached.
• Fix: Gemini/Vertex backends now throw NotSupportedException when WithMaxIterations is used instead of silently ignoring the value.

MaIN.NET.0.10.12

0.10.10 release

• Adds WithMaxIterations() to ToolsConfigurationBuilder and McpContext to override the default tool-call iteration limit.
• Fix: MCP loop now sends a final synthesis request instead of returning an error string when the iteration cap is reached.
• Fix: Gemini/Vertex backends now throw NotSupportedException when WithMaxIterations is used instead of silently ignoring the value.

MaIN.NET.0.10.11

0.10.11 release

• Adds WithMaxIterations() to ToolsConfigurationBuilder and McpContext to override the default tool-call iteration limit.
• Fix: MCP loop now sends a final synthesis request instead of returning an error string when the iteration cap is reached.
• Fix: Gemini/Vertex backends now throw NotSupportedException when WithMaxIterations is used instead of silently ignoring the value.

MaIN.NET.0.10.10

0.10.10 release

• Adds WithMaxIterations() to ToolsConfigurationBuilder and McpContext to override the default tool-call iteration limit.
• Fix: MCP loop now sends a final synthesis request instead of returning an error string when the iteration cap is reached.
• Fix: Gemini/Vertex backends now throw NotSupportedException when WithMaxIterations is used instead of silently ignoring the value.

MaIN.NET.0.10.9

0.10.9 release

• Expands model catalog with latest Claude (Opus 4.7, Sonnet 4.6, Haiku 4.5), OpenAI (GPT-4.1, GPT-5 family, o3, o4-mini, Codex Mini), Gemini (3.x series), Groq (Llama 4, Qwen3, Compound), and Ollama (Llama 4, Gemma 3/4, Qwen3/3.5/3.6/Coder, DeepSeek R1, Phi4, Mistral) models.
• Adds configurable Ollama base URL (MaIN__OllamaBaseUrl) for containerized deployments.
• Adds InferPage Docker images published to GHCR with CPU, CUDA, and bundled-Ollama targets.

MaIN.NET.0.10.5

Adds agent skills support.

What's Changed

• feat: Skills support by @Madzionator in #131

Full Changelog: 0.10.4...0.10.5

MaIN.NET.0.10.4

0.10.4 release

Adds Google Vertex AI as a backend with authentication, MCP support, and new models including image generation, along with UI configuration and example usage.

MaIN.NET.0.10.3

0.10.3 release

Inference parameters are now backend-specific — each AI provider has its own typed params class where only explicitly set values are sent to the API, with an AdditionalParams dictionary for custom fields.