Skip to content

Latest commit

 

History

History
114 lines (77 loc) · 2.22 KB

README.md

File metadata and controls

114 lines (77 loc) · 2.22 KB
Raw

GenerativeSQL App

This system aims to generate sql queries based in the context of a question or instruction.

Functionalities

  • Receives a text input and checks if the input is valid.
  • If the input is in Portuguese, it translates it to English.
  • With the input processed, the Vanna AI service is used to generate a SQL query.

Requirements

It's necessary to create an account on the Vanna AI platform and also obtain a key for Google Gemini.

Vanna AI (https://vanna.ai/)

Google Gemini (https://ai.google.dev/aistudio)

In Vanna platform, create a new model, which will be used in the configuration (VANNA_MODEL).

Technologies

  • Python 3.11 with FastAPI
  • MySQL 8.0.32

Quickstart - Docker and Tests

First time

Create a build/mysql folder, to hold a volume for the database.

$ mkdir -t build/mysql
$ chmod -R 777 build/mysql

Create a copy of .env.example and fill the variables with the values of your choice (i.e. your database credentials, etc)

$ cp .env.example .env

Make shell script files executable

$ chmod +x start.sh

Run at project root

$ ./start.sh

Run the tests to verify that the installation completed successfully.

$ docker exec -it api pytest
$ ./commit.sh

Running after build

$ docker compose up -d

Create tables in an Access database (named "northwind_test" as specified in your .env configuration file). This is ideal for local testing.

$ docker exec -it api mysql -u root -p -D northwind_test

After access the docker with above command, at the MySQL prompt type:

$ source /tmp/scripts/schema.sql
$ source /tmp/scripts/data.sql

After setting up the database connection, run the following script:

$ docker exec -it api python /api/scripts/vanna_mysql_training.py

Testing via CLI

$ docker exec -it api python scripts/vanna_mysql_question.py \
    "What are the top 10 customers by sales ? (Include the customer's full name)"

Access documentation

In the browser, access http://0.0.0.0:8003/docs, for access to API methods in OpenAPI format.

Running tests

$ docker exec -it api pytest