LiteAgent

KO | EN

gitlite — search

#javascript #python #hacktoberfest #llm #go #golang #android #machine-learning #nodejs #windows #awesome #kubernetes #awesome-list #ios

(★ 27)

AI Agents for Industrial Applications 工业领域的AI智能体

LiteAgent Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# LiteAgent

English · [中文](README-zh_CN.md)

LiteAgent is an open-source AI Agent platform designed for industrial applications, enabling users to quickly build and deploy AI Agent applications. It supports integration with various models and tools, providing web, desktop, and backend services suitable for diverse industrial scenarios.

## Program description
  - [Web](lite_agent_web/README.md): manage agent, tool, model, knowledge base and user, etc.
  - [Desktop](lite_agent_client/README.md): manage agent, tool, model, etc.
  - [Backend](lite_agent_backend/README.md): service and api for web, desktop and sdk.
  - [Plugin Engine](lite_agent_plugin_runner/README.md): Runs plugins, such as WeChat and DingTalk bots
  - [SDK](https://github.com/LiteVar/LiteAgent/tree/master/lite_agent_sdk): C#/LabVIEW/TypeScript/Dart/Java sdk, to help developers quickly integrate LiteAgent.

## Features

### 1. Agent

- Ordinary text chat agent
- Tool agent
- Multi-agent
- General agent(Web)
- Distribution agent(Web, Desktop)
- Reflection agent(Web, Desktop)
- Agent mode support: Parallel, Serial, Reject
- Auto Multi Agent (Web, Desktop)
- Support exporting/importing configurations
- Support configuring and retaining recent conversation rounds (default: 5 rounds)
- Chat supports image and video uploads

### 2. Tool Support

- Protocol: openapi, openrpc, openmodbus
- Tool mode support: Parallel, Serial, Reject
- Tool Sequence: Reflection on Tool Order
- Supports third-party OpenTool callbacks
- Protocol: MCP
  - Server support MCP over SSE
  - Desktop support MCP over stdio
- Protocol: OpenTool
  - Support for OpenTool-formatted specification documents
  - Support calling OpenTool Server
  - Support StreamCall invocation mode
- Support exporting tools

### 3. Functional Modules

- LLM management(Web, Desktop)
- Tool management(Web, Desktop)
- Agent management(Web, Desktop)
- User management(Web)
- Knowledge base management(Web)
- Added reasoning type differentiation
- Preview mode added for prompt and knowledge base document editors
- New SDKs released (C#/LabVIEW/TypeScript/Dart/Java)
- System-level model management (Web)
- System-level plugin management (Web)

### 4. LLM support

- Only support OpenAI Style API，other models(like: Zhipu-AI、QianFan、ChatGLM、Kimi、qwen and so on), please use the OneAPI tool to convert to OpenAI style.
- Speech-to-Text (ASR model), supporting real-time speech-text streaming conversion
- Text-to-Speech (TTS model), supporting real-time text-speech streaming conversion
- LLM supports exporting/importing configurations in JSON format
- Support adding system-level large models for all system users
- Support adding VL models (Vision-Language Models)

### 5. Knowledge Base Support

- Convert documents to markdown, including text and images (Web)
- Add document content summarization (Web)
- Knowledge base retrieval, support "View Original" and "Download Original Document" (Web, Desktop)
- Knowledge base list, can select and "Download as Markdown" (Web, Desktop)

### 6. Plugin Smart Integration
- Provide CRUD functionalities
- Enable Agent integration with WeChat (MP), DingTalk (Robot), and Website platforms to bring intelligent capabilities to the platforms
- Support multi-dimensional session data statistics and trend analysis

### 7. ChatKit Component
- Support component-based custom chat window configuration
- Easily integrate intelligent agents into homepages/websites

## VDB migrate to Qdrant
- Please refer to [lite_agent_backend/docs/milvus-qdrant-migrate/README.md](lite_agent_backend/docs/milvus-qdrant-migrate/README.md)

## Quick Start

### 1. Minimum System
Before using LiteAgent, make sure your machine meets the following minimum system requirements:
 
>- CPU >= 4 Core
>- RAM >= 10 GiB
>- Hard disk >= 256GB(Recommended SSD)

### 2. Invitation Email Configuration

- Configure invitation email addresses and email accounts for creating and inviting users
- Edit `docker/LiteAgent/config/application.yml`, change host, username, password and port
```
mail:
  host: smtp.xxx.com
  username: liteagent@xxx.com
  password: XXXYYYZZZ
  port: 465
```

### 3. Run
#### 3.1 Preparation
> - Build LiteAgent web in nodejs environment(please refer to lite_agent_web/README.md), and then replace it to the directory: docker/nginx/html/
> - Build LiteAgent backend app in java environment(please refer to lite_agent_backend/README.md),
>   - rename to lite-agent.jar, and then replace it to the directory: docker/LiteAgent/app/
>   - update config file to the directory: docker/LiteAgent/app/config (please refer to lite_agent_backend/lite-agent-rest/src/main/resources)

#### 3.2 Beginning
- Pull up services by docker compose
- Please make sure docker, docker compose plugin are installed on your machine
```
cd docker 
docker compose up -d
```

### 4. Init Root User

1. Open the browser and access the initialization page( [http://<YOUR_IP>:8080/init](http://<YOUR_IP>:8080/init) )
2. Enter email address, nickname, password, and confirm password, click login to create the root user
3. Move to the top of the avatar and click on "Manage My Workspace"
4. Then you can manage agents, tools, models and users
   - When creating an agent, adding a model is necessary and tools are optional  
5. After creating the agent, you can start chatting

### 5. New User Login

- Access the login page([http://<YOUR_IP>:8080](http://<YOUR_IP>:8080) )

LiteAgent

LiteAgent Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

LiteAgent

LiteAgent Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README