FastGPT README

The project's README from github.com/labring/FastGPT. As one markdown file: README.md.

fastgpt logo

FastGPT

English | 简体中文 | Bahasa Indonesia | ไทย | Tiếng Việt | 日本語

FastGPT 是一个 AI Agent 构建平台,提供开箱即用的数据处理、模型调用等能力,同时可以通过 Flow 可视化进行工作流编排,从而实现复杂的应用场景。

cloud document development project

https://github.com/labring/FastGPT/assets/15308462/7d3a38df-eb0e-4388-9250-2409bd33f6d4

快速开始

可以通过 Docker 快速启动 FastGPT,在终端输入以下命令,根据引导完成输入即可拉取配置。

# 输入命令拉取配置文件
bash <(curl -fsSL https://doc.fastgpt.io/deploy/install.sh)
# 启动
docker compose up -d

完全启动后,可通过 http://localhost:3000 访问 FastGPT,默认账号为 root,密码为 1234。

如果你遇到问题,可以查看完整 Docker 部署教程

🛸 使用方式

💡 核心功能

Demo Demo
Demo Demo

1 应用编排能力

2 应用调试能力

3 知识库能力

4 插件能力

5 运营能力

#

💪 我们的项目 & 链接

#

🌿 第三方生态

#

🏘️ 社区交流群

扫码加入飞书话题群:

#

🤝 贡献者

我们非常欢迎各种形式的贡献。如果你对贡献代码感兴趣,可以查看我们的 GitHub Issues,大展身手,向我们展示你的奇思妙想。




Active participants of labring - past 28 days **** New trends of labring
New participants of labring - past 28 days

🌟 Star History

Star History Chart #

使用协议

本仓库遵循 FastGPT Open Source License 开源协议。

  1. 允许作为后台服务直接商用,但不允许提供 SaaS 服务。
  2. 未经商业授权,任何形式的商用服务均需保留相关版权信息。
  3. 完整请查看 FastGPT Open Source License
  4. 联系方式:[email protected],点击查看商业版定价策略

FastGPT (labring/FastGPT)

Searching Database…

29.8K ·GitHub
What is this?FastGPT is an AI agent building platform: a knowledge-based platform on top of LLMs with out-of-the-box data processing, model calling, RAG retrieval and visual workflow orchestration, aimed at building complex question-answering systems without heavy setup.
Is this a hosted service, or something I run on my own server?Both: there is a cloud version at fastgpt.io, a community self-hosted edition you deploy with Docker or one-click on Sealos Cloud, and a commercial edition with more features and hands-on support.
How do I get it running?Start with Docker: run the quick-start command from the README and follow the terminal prompts to pull the config. When it is fully up, open the web UI and log in with the default account root and password 1234.
What do I need to hack on the source code?Node.js 22.23.2 or newer and pnpm: it is a Turbo-managed monorepo, dev scripts run through turbo, and a Chakra UI theme codegen step executes automatically on install.
Do I need to be a developer to build apps with it?Building itself is low-code: you compose chat workflows, plugin workflows (with basic RPA nodes), user interaction and agent skills on a visual Flow canvas. Running your own instance still means a Docker deployment.
Which AI models can it plug into?It calls models through configurable providers — the repo lists OpenAI, Claude, DeepSeek and Qwen — and the team also maintains AI Proxy, a companion model aggregation and load-balancing service.
What can I load into its knowledge base?TXT, MD, HTML, PDF, Docx, PPTX, CSV and XLSX files, plus URL reading, CSV batch import, manual input, direct segmentation and QA-split import. Libraries can be reused and mixed, chunks edited or deleted, retrieval is hybrid with rerank, and API-based knowledge bases are supported.
How do I debug an app I have built?You can test knowledge base searches in isolation, act on citation feedback users leave in chats, inspect full call-chain logs and run app evaluations. More advanced orchestration debugging is still on the roadmap.
Can I share or embed the apps I build?Yes: apps can be shared through a login-free window or embedded anywhere with one-click iframe embedding. Operators also get centralized chat history with data annotation and app operation logs.
Creator
labring
Size
459.1 MB
Stars
29.8K
Age
3y
Commits · 90 Days