Repository files navigation

A powerful AI agent application.Still under development; pull requests are welcome.

here is neroi stack wiki

• MCP-Plugin architecture for extensibility
• Integration with OpenAI and other AI services
• Support for Multi-Agent Orchestration

• Expand plugin ecosystem for third-party integrations
• Optimize performance for large-scale AI tasks
• agentic workflows and collaboration features
• STT-TTS support

1. Ensure you have.NET 10 SDK installed. You can download it from the.NET website.

1. Clone the repository, and open it in your IDE

   git clone https://github.com/tse-wei-chen/NeroiStack.git

2. If you use Visual Studio Code, just PressF5 hard.

3. Build the project using.NET CLI

   dotnet build

4. Run the application

   dotnet run --project NeroiStack/NeroiStack.csproj

NeroiStack supports various multi-agent orchestration strategies to enable complex AI workflows. Below are some of the key strategies illustrated with diagrams:

1. Concurrent

   graph TDInput[Input] --> Agent1[Agent 1]Input[Input] --> Agent2[Agent 2]Input[Input] --> Agent3[Agent 3]Agent1[Agent 1] --> C[Collector]Agent2[Agent 2] --> C[Collector]Agent3[Agent 3] --> C[Collector]C[Collector] --> Output[Output]

   Loading
2. Sequential

   graph TDInput[Input] --> Agent1[Agent 1]Agent1[Agent 1] --> Agent2[Agent 2]Agent2[Agent 2] --> Agent3[Agent 3]Agent3[Agent 3] --> Output[Output]

   Loading
3. Group Chat

   graph TDInput[Input] --> GroupChatManager[GroupChatManager]GroupChatManager[GroupChatManager] <--> Agent1[Agent 1]GroupChatManager[GroupChatManager] <--> Agent2[Agent 2]GroupChatManager[GroupChatManager] <--> Agent3[Agent 3]GroupChatManager[GroupChatManager] --> Output[Output]

   Loading
4. Handoff

   graph TDInput[Input] --> A1[Entry Agent]A1 -- Transfer --> A2[Specialized Agent A]A1 -- Transfer --> A3[Specialized Agent B]A2 -- Complete --> Output[Output]A3 -- Complete --> Output[Output]

   Loading
5. Magentic

   graph TDInput[Input] --> M[Magentic Manager]M -- Select & Call --> A1[Agent 1]A1 -- Observation --> MM -- Select & Call --> A2[Agent 2]A2 -- Observation --> MM -- "Satisfied (Done)" --> Output[Output]

   Loading