Skip to main content

Chat with your existing database without using vector DB.

Project description

DBChat

An agentic AI tool designed to seamlessly interact with existing databases using an agentic AI approach. This project aims to simplify the process of querying databases, making it more intuitive and accessible for everyone, regardless of their technical background.

Features

  • Uses an Agentic approach to answer to user's question
  • Safe from user executing malicious DDL queries.
  • Can add Data-safety by only exposing certain tables and columns in the schema.
  • Doesn't expect the LLM to generate executable code, rather behaves as an reasoning engine
  • reduced hallucinations and more deterministic in nature
  • no need for vector DBs
  • Supports all relational databases supported by sql alchemy. Refer https://docs.sqlalchemy.org/en/20/dialects/

How to Run?

docker pull dhanilan/dbchat
docker run -p 5173:5173 -p 8000:8000 dhanilan/dbchat

default database to save connections and conversation is mongo at mongodb://localhost:27017/dbchat

if you want to save them at specific mongo database pass it as an env

docker run  -p 5173:5173 -p 8000:8000 -e DB_URL=mongodb://host.docker.internal:27017/dbchat dhanilan/dbchat

if you want to run spider dataset from https://yale-lily.github.io/spider

docker run  -p 5173:5173 -p 8000:8000 -e DB_URL=mongodb://host.docker.internal:27017/dbchat -e ATTACH_SPIDER_DATASET=1 dhanilan/dbchat

using docker-compose

The docker compose also comes with a chinook database for testing. You can use it by adding a connection to the chinook database in the UI use postgresql+psycopg2://postgres:postgres@chinook:5143/chinook as the connection string

update the docker-compose.yml with the required envs if neccessary and run

docker-compose up -d

Architecture

Built with Autogen.

alt text

Development

Dev container contains all the necessary deps , mongodb storing the schema and chat history , conversations, settings etc and also a chinook database in postgres for playing around

Installing dependencies

UI

pip install -r requirements.txt

server

cd src/ui && npm i

To run the Application in local

UI

cd src/ui && npm run dev

API

cd src && uvicorn api.main:app

Open the UI

By default opens in http://localhost:5173/

Goto setting to save the OPEN AI API key

connect to dev container chinook db

postgresql+psycopg2://postgres:postgres@localhost:5432/chinook

Roadmap

  • clean up some code for now
  • Add more tests
  • add feature to annotate the schema
  • add few shots of the conversation to the schema
  • add feature to add custom prompts
  • add access control for tables and columns `

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

py_dbchat-0.0.2rc126.post1.tar.gz (629.7 kB view details)

Uploaded Source

Built Distribution

py_dbchat-0.0.2rc126.post1-py3-none-any.whl (21.4 kB view details)

Uploaded Python 3

File details

Details for the file py_dbchat-0.0.2rc126.post1.tar.gz.

File metadata

  • Download URL: py_dbchat-0.0.2rc126.post1.tar.gz
  • Upload date:
  • Size: 629.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: python-requests/2.32.3

File hashes

Hashes for py_dbchat-0.0.2rc126.post1.tar.gz
Algorithm Hash digest
SHA256 ede667b6797c953b606eb25e91ca104c0cc70d757f0c2b3a63ecea0f71c306ce
MD5 ce30abf4c707f55c38f25a314c943aa3
BLAKE2b-256 58bf5a6da4b24928dad261e91d6ee8aa50d651899ca73d87d787ccd270c01e9a

See more details on using hashes here.

File details

Details for the file py_dbchat-0.0.2rc126.post1-py3-none-any.whl.

File metadata

File hashes

Hashes for py_dbchat-0.0.2rc126.post1-py3-none-any.whl
Algorithm Hash digest
SHA256 f0539313dcfba4361fc3fee4c5f9a67ce245007c5ede365fca6b36980febcc31
MD5 d39363111ee7e9ccf2b1643bb309a82f
BLAKE2b-256 a4ccdc1a0e82df9029eecf6a4a77a7cfce0abd1df46cc4fa4845a0d7d1a94796

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page