/stock-chat-flowise

An LLM-backed chatbot regarding stocks, implemented by Flowise and Langchain JS.

MIT LicenseMIT

stock-chat-flowise

An LLM-backed chatbot regarding stocks, implemented by Flowise and Langchain JS.

How to make an LLM-backed app in minutes?

This repo is about building an LLM-backed app with low-code frameworks like Flowise (written in typescript) and Langflow (in python), from the perspective of absolute beginners.

Flowise experiment

Step 0 - Get an openAI api (or other supported api)

Option 1 - Local setup & development

For mac users with anaconda for package management, check node and npm version first. Make sure the node version is above 18.15.0. In the terminal:

node --version 
npm -version

Next, install Flowise and start it via your local port

npm install -g flowise

npx flowise start

Next - Start building your app locally as LEGO game

Open http://localhost:3000 and you'll see a workbench with all components from LangChain! Start this LEGO-like coding game and build what you want! 😄 You can load the json file from an interested project and load it into your workbench so to replicate anyone's effort in a minute.

Option 2 - Deploy Flowise to the cloud (recommended)

The official document provides several ways to deploy your app.

  • Render, Railway, Replit.
    • Easier to implement but the data will be lost once the app restarts.
  • AWS, DigitalOcean, GCP.
    • More reliable but can be technically challenging.

Here we'll test the render option. Follow the instruction of document and the reference video as below.

After deployment, the rest of LEGO coding is similar as local development.

Note - App can be instantly used as API or embed in website 😍

See the screenshots in this project folder.

文科小白都能立即上手的大模型应用框架！

非常欢迎参与本开源项目的贡献！

补充项目背景

首先感谢开源社区 DataWhale的用心组织~ 让纯文科出身的Sarai同学也萌发了搭建大模型应用的热情。

项目最初目标是参考Finchat等应用搭建具备本地PDF总结问答、web搜索总结等功能的股票分析助手，可选用多种LLM模型或商业API。

社区课程中推荐的LangChain是一个强大的框架，其提供了一套工具、组件和接口，可简化创建由大型语言模型 (LLM) 支持的app开发过程，并集成额外的资源，例如 API 和数据库等。

虽然Langchain已经很方便了，但对于非程序员来说还是有些麻烦（和恐惧）。对于不熟悉代码和或没时间折腾各种框架的小伙伴来说，通过UI直观拖拽组件、输入关键参数就能实现零代码开发专属gpt应用的开源框架简直就是天赐神器！

我们重点探索尝试了基于Langchain框架的开源项目Flowise和Langflow。 其中Flowise是typescript写的，Langflow是python写的，基本使用颇为相似。 这里着重介绍了Flowise的使用。

补充学习资料

国内这方面的资源貌似还不是很多，资料请参考：

开发计划（暂定）

后续在此基础上进一步探索基于Langchain的低代码框架的应用潜力和优缺点。 同时类似功能使用Langchain + Streamlit等框架重写并对比效果。 期待大佬去这里 GitHub - AldeaTeam/stock-chat-gpt 贡献代码~~

致谢

感谢本组成员的积极参与和献言献策~ 特别感谢： KuangJunWei、 EricJiang,BlankZhou, DoubleShan当然以及杨老师的鼓励指导~~ 组长Sarai小白