Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/cogitovirus/langchain-sql-agent-bootstrap
Langchain SQL Agent Boostrap application. Flask on the backend, React on the front.
https://github.com/cogitovirus/langchain-sql-agent-bootstrap
flask langchain openai sql
Last synced: 3 months ago
JSON representation
Langchain SQL Agent Boostrap application. Flask on the backend, React on the front.
- Host: GitHub
- URL: https://github.com/cogitovirus/langchain-sql-agent-bootstrap
- Owner: cogitovirus
- License: mit
- Created: 2023-03-24T17:56:12.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2023-03-31T23:24:54.000Z (over 1 year ago)
- Last Synced: 2024-07-13T02:31:51.660Z (4 months ago)
- Topics: flask, langchain, openai, sql
- Language: Python
- Homepage: https://cogitovirus.com/posts/20230402-gpt-4-langchain-agent-bootstrap/
- Size: 2.22 MB
- Stars: 31
- Watchers: 3
- Forks: 6
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- jimsghstars - cogitovirus/langchain-sql-agent-bootstrap - Langchain SQL Agent Boostrap application. Flask on the backend, React on the front. (Python)
README
# Langchain SQL Agent Bootstap
This is a simple App for testing LLM to SQL commands on a sqlite database using [Langchain SQL Agent](https://python.langchain.com/en/latest/modules/agents/toolkits/examples/sql_database.html). The repo comes with a setup script that loads a sqlite database with some sample data.
> This repo is intended to be a starter for Langchain and also a starter for programming with co-pilot GPT-4. The scope is small enough that you can even get by with copying/pasting whole clases to GPT-4 and asking for explanations or improvemnts to the code.
>There's a single important TODO item that is not and will not be ever completed, which is to implement the file upload functionality in the UI. The backend is already setup to handle file uploads (csv). If you're interested you could probably also:
> - clean the ouput of the agent after each execution
> - add a auto-refresh or a refresh button to the tables
Backend has several endpoints:
- `/api/v1/tables` - returns a list of tables in the database
- `/api/v1/tables/` - returns a list of columns in the table
- `/api/v1/run-command` - takes a json payload with the command to execute and returns the results via a websocket
- `/api/v1/upload` - takes a csv file and loads it into the database
`run-command` is where all the fun is happening with the langchain sql agent.
## Running locally
### Pre-requisites
Backend was setup with Python 3.10.10
Frontend was setup with Node 18.15.0
Experience may vary with other versions
You have to have an open-api-api-key to use the langchain sql agent.
Copy the `.env.example` file to `.env` and add your key to the `OPENAI_API_KEY` variable. You can also change the model which is used to make predictions. In the time of writing, `gpt-4` is in closed beta, so the backend might not work if you select that model.
### Backend
Create virtual environment (only need to do this once)
```bash
python3 -m venv data-whisperer-backend/myenv
```
activate virtual environment
```bash
source data-whisperer-backend/myenv/bin/activate
# install dependencies
pip install -r data-whisperer-backend/requirements.txt
```
**seed the database**
```bash
python data-whisperer-backend/setup/create_and_seed_db.py
```
finally, run the backend
```bash
python data-whisperer-backend/run.py
```
### Frontend
```bash
cd data-whisperer-frontend/
npm install # only need to do this once
npm start
```
## Teardown
```bash
# deactivate virtual environment after you are done
deactivate
```