# langchat
**Repository Path**: tree23/langchat
## Basic Information
- **Project Name**: langchat
- **Description**: LangChat 是基于Java SpringBoot3 和 Vue3开源的企业级 AI Agent 应用平台，支持多模型、Agent、知识库 RAG、Skills、MCP 与智能问数。
- **Primary Language**: Unknown
- **License**: Apache-2.0
- **Default Branch**: next
- **Homepage**: http://langchat.cn
- **GVP Project**: No
## Statistics
- **Stars**: 0
- **Forks**: 690
- **Created**: 2026-10-29
- **Last Updated**: 2026-10-29
## Categories & Tags
**Categories**: Uncategorized
**Tags**: None
## README
## LangChat Agentic enterprise edition
[LangChat Agentic](https://agentic.langchat.cn/) is the commercial product from LangChat Team for enterprise procurement, delivery, and long-term operations. It provides a broader and more complete capability set than the current open-source edition and is designed for organizations deploying AI applications across real business workflows.
The enterprise edition includes:
- standard agents, workflow agents, and agentic applications;
- long-term memory, memory compression, isolated session workspaces, and containerized sandboxes;
- Supervisor, sequential, parallel, and loop-based multi-agent orchestration;
- enterprise knowledge, hybrid retrieval, versioned indexes, result fusion, and optional reranking;
- governed Text-to-SQL analytics, enterprise data connectivity, and explainable charts;
- Function Call, MCP, A2A, Skill, OpenAPI, and HTTP extension channels;
- model governance, access control, content safety, usage limits, cost control, and runtime auditing;
- private deployment, localized infrastructure support, implementation, training, and ongoing releases.
For enterprise AI platform procurement or vendor evaluation, we recommend reviewing LangChat Agentic first:
[Explore the enterprise edition](https://agentic.langchat.cn/) · [Read the product brief](https://agentic.langchat.cn/docs) · [View pricing and delivery](https://agentic.langchat.cn/price) · [Meet the team](https://agentic.langchat.cn/about)
## About LangChat
LangChat is a modular, extensible, open-source platform for building AI agent applications. It brings model connectivity, agent building, knowledge-base RAG, Skills, MCP, data analysis, and access control into one Java and Vue stack.
The open-source edition is designed for technical evaluation, custom development, and internal team use. Its modular monolith architecture keeps deployment straightforward while preserving clear boundaries for future evolution.
## Capabilities
| Capability | Description |
| --- | --- |
| Model connectivity | Manage OpenAI, Azure OpenAI, Anthropic, Gemini, DeepSeek, DashScope, Zhipu AI, Ollama, and OpenAI-compatible services |
| Agent builder | Configure models, system prompts, knowledge bases, Skills, and MCP services with dedicated build and runtime views |
| Chat runtime | Manage conversations and messages, stream responses over SSE, and expose an OpenAI-compatible Chat Completions API |
| Knowledge-base RAG | Manage knowledge bases, document uploads, parsing previews, chunks, index status, and vector-store settings |
| Skills | Upload skill packages or folders, browse files, download packages, and register runtime tools |
| MCP integration | Manage external MCP servers and reuse client sessions during agent execution |
| Conversational analytics | Connect data sources, inspect schemas, and return natural-language analysis with charts |
| Image AI | Generate and recognize images through the shared model configuration system |
| Application marketplace | Publish, discover, and run agents from a unified chat experience |
| Access control | Manage authentication, users, roles, menus, dynamic routes, and RBAC permissions |
## Highlights
- **Unified AI runtime:** clear runtime definitions connect models, agents, knowledge, Skills, and MCP without spreading provider-specific SDK code through the business layer.
- **Open integration surface:** platform APIs and an OpenAI-compatible endpoint make existing clients and internal systems easier to connect.
- **Extensible tool system:** skill files, runtime tool registration, and MCP services can be composed per agent.
- **Knowledge and data in one product:** document retrieval and structured-data analysis share one application platform.
- **Modular monolith:** common, auth, aigc, core, datasource, and server modules balance development speed with explicit domain boundaries.
- **Complete management UI:** the Vue 3 workspace covers models, agents, knowledge, Skills, MCP, data sources, users, roles, and core operational flows.
## Knowledge-base interface
| Document management |
Document segments |
 |
 |
## Architecture
```mermaid
flowchart LR
U[Users / external systems] --> UI[Vue 3 console]
U --> API[REST / OpenAI-compatible API]
UI --> API
API --> AUTH[langchat-auth
Authentication and RBAC]
API --> AIGC[langchat-aigc
Configuration and resources]
API --> CORE[langchat-core
Agent / Chat / RAG / Skills]
API --> DS[langchat-datasource
Data sources and analytics]
CORE --> LLM[Model providers
OpenAI / Gemini / Ollama / ...]
CORE --> MCP[MCP services and tools]
CORE --> VECTOR[Pgvector / Milvus]
AIGC --> MYSQL[(MySQL)]
AUTH --> MYSQL
DS --> DATA[(Business data sources)]
```
### Technology stack
| Layer | Technologies |
| --- | --- |
| Backend | Java 17, Spring Boot 3.3, LangChain4j 1.12, MyBatis-Plus, Sa-Token |
| Frontend | Vue 3, TypeScript, Vite, Pinia, Naive UI, Tailwind CSS, VXE Table, ECharts |
| Data | MySQL, PostgreSQL / Pgvector, Milvus |
| Engineering | Maven multi-module build, pnpm workspace, SSE, OpenAI-compatible API |
### Repository layout
```text
langchat
├── langchat-common # Shared foundations and dependency management
├── langchat-auth # Authentication, users, roles, menus, and permissions
├── langchat-aigc # Models, agents, knowledge bases, and Skills
├── langchat-core # Chat runtime, RAG, MCP, images, and data analysis
├── langchat-datasource # Data-source connectivity and schema access
├── langchat-server # Spring Boot entry point and configuration
├── langchat-ui # Vue 3 frontend workspace
└── docs/db # Database initialization scripts
```
## Quick start
### Requirements
- JDK 17+
- Maven 3.9+
- Node.js 22.22+
- pnpm 10.32+
- MySQL 8+
### 1. Clone the repository
```bash
git clone -b next https://github.com/langchat/langchat.git
cd langchat
```
### 2. Configure the local environment
The local configuration file is ignored by Git:
```bash
cp langchat-server/src/main/resources/application-local.example.yml \
langchat-server/src/main/resources/application-local.yml
```
Edit `application-local.yml` and provide the database connection and local administrator password. Use environment variables in production and never commit credentials.
Initialize the database with `docs/db/langchat.sql`.
### 3. Install frontend dependencies
```bash
cd langchat-ui
corepack enable
pnpm install
cd ..
```
### 4. Start LangChat
```bash
SPRING_PROFILES_ACTIVE=local ./langchat.sh
```
Default endpoints:
- Frontend: `http://localhost:5888`
- Backend: `http://localhost:8080`
- Health check: `http://localhost:8080/actuator/health`
To run each part separately:
```bash
mvn -pl langchat-server -am spring-boot:run
pnpm --dir langchat-ui --filter @vben/web-naive dev
```
## Configuration and security
- `application.yml` contains shared settings.
- `application-dev.yml`, `application-test.yml`, and `application-prod.yml` define environment-specific settings.
- `application-local.yml` contains machine-local values and is ignored by Git.
- Production values are injected through `LANGCHAT_MYSQL_*`, `LANGCHAT_ADMIN_*`, and related environment variables.
- Never disclose API keys, database passwords, or access tokens in issues, logs, commits, or screenshots.
## Contributing
Issues, design discussions, and pull requests are welcome. Read the [contribution guide](CONTRIBUTING.md) and [code of conduct](CODE_OF_CONDUCT.md) before contributing. Report vulnerabilities privately according to the [security policy](SECURITY.md).
The active development baseline is the `next` branch. Create feature branches from `next` and open pull requests against `next`.
## Commercial support
LangChat Agentic is the complete commercial product for enterprise procurement. Organizations planning an enterprise agent platform, private deployment, systems integration, implementation, or ongoing support should evaluate [LangChat Agentic](https://agentic.langchat.cn/) directly.
- [LangChat website](https://langchat.cn)
- [LangChat Agentic](https://agentic.langchat.cn/)
- [Product brief](https://agentic.langchat.cn/docs)
- [Pricing and delivery](https://agentic.langchat.cn/price)
## License
Unless a subdirectory states otherwise, this project is licensed under the [Apache License 2.0](LICENSE). `langchat-ui` includes Vue Vben Admin-derived code distributed under the MIT License. See [langchat-ui/LICENSE](langchat-ui/LICENSE) and [NOTICE](NOTICE).
Copyright © 2026 LangChat Team. 
