MaIN.NET.0.7.2

0.7.2 release

• allow any Stream to be used as fileinfo
• Chat/Agent context extension (accept multiple messages)

MaIN.NET.0.7.1

0.7.1 release

• Chat/Agent context extension (accept multiple messages)

MaIN.NET.0.7.0

0.7.0 release

• Add full tool support for all available cloud providers

On chat/agent context

.WithTools(new ToolsConfigurationBuilder()
                .AddTool(
                    name: "get_current_time",
                    description: "Get the current date and time",
                    execute: Tools.GetCurrentTime) 
                .WithToolChoice("auto")
                .Build())

MaIN.NET.0.6.4

0.6.4 release

Grammar handling has been refactored to support multiple grammar formats. Definitions must now be wrapped in a Grammar object with an explicit format selection. Currently supported formats: GBNF, JSONSchema.

// before:
.WithInferenceParams(new InferenceParams
{
Grammar = grammarText;
})

// after:
.WithInferenceParams(new InferenceParams
{
Grammar = new Grammar(grammarText, GrammarFormat.GBNF)
})

MaIN.NET.0.6.3

0.6.3 release

• Allow vision message processing for cloud LLM providers

MaIN.NET.0.6.2

0.6.2 release

• Allow vision message processing for cloud LLM providers

MaIN.NET.0.6.1

0.6.1 gemini imagen model naming fix

0.6.0 release

• WithCustom model fix to use correct path
• Update packages to newest versions
• Support custom RAG implementation for deepseek & groq
• New local models added

Models:

• NomicV2 (Embedding)
• Lfm2-1.2b
• Minicpm4-8b
• Webgen-4b
• Gemma3n-e4b
• Mistral3.2-24b
• Bielik2.5-11b

MaIN.NET.0.6.0

0.6.0 release

• WithCustom model fix to use correct path
• Update packages to newest versions
• Support custom RAG implementation for deepseek & groq
• New local models added

CLI: https://maindoc.link/#/doc/cli

Models:

• NomicV2 (Embedding)
• Lfm2-1.2b
• Minicpm4-8b
• Webgen-4b
• Gemma3n-e4b
• Mistral3.2-24b
• Bielik2.5-11b

MaIN.NET.0.5.1

0.5.1 release

Anthropic integration has been added.

MaIN.NET.0.5.0

0.5.0 Release

• Knowledge feature including 3 examples with web source, file source, and MCP source.

Knowledge Database

This feature enables the creation of agents that can provide impressive data sourcing with minimal context. This is achieved by persisting knowledge as an index.json file, which contains a set of sources with assigned tasks. All prompts for agents using this feature can verify if there is a matching tag in the knowledge memory and use only that information to answer questions. To understand how it works, the best way is to check out the examples.
Documentation: https://maindoc.link