Dataherald is a Python-based natural language to SQL engine that exposes an API to query relational data. It supports multiple database backends and features finetuning capabilities, with releases adding Redshift, Astra vector store, ClickHouse, and MariaDB/finetuning related updates.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Dataherald is a natural language-to-SQL engine designed to answer questions over relational data by exposing an API connected to a database. It includes an engine, enterprise API layer, admin console, and a Slackbot component as part of a monorepo. The project targets enterprise-style deployment with authentication and observability considerations described across its components.
How it works
The repository contains multiple services under /services (Engine, Enterprise, Admin-console, Slackbot). The README indicates a docker-based local running pattern where each component has its own docker-compose.yml, and a root script (docker-run.sh) to start all services on a common Docker network. Environment variables are per-service (see .env.example in each service directory). The major use case is natural language input translated to SQL to query a relational database via an API.
Getting started
Getting started guidance is in the README under Running locally:
- Each component in /services has its own docker-compose.yml.
- Create a .env file for each service using the corresponding .env.example.
- For the Next.js front-end app, use .env.local.
- Run all services with:
sh docker-run.sh
Recent releases
Latest releases include:
- 1.0.3 v1.0.3 (2024-04-30): Added Redshift support; improved features listed as "New features".
- 1.0.2 v1.0.2 (2024-04-03): Adds Astra vector store support.
- 1.0.1 v1.0.1 (2024-03-04): Added ClickHouse support.
- 1.0.0 v01.0.0 (2024-01-16): Resource and Endpoints updates including Finetuning notes.
- 0.0.6 v0.0.6 (2023-11-13): Changes to POST /api/v1/responses endpoint behavior.
Traction
- Stars: 3641
- Forks: 262
- Open issues: 21
Behind the repo
- The project is a monorepo with four components under /services: Engine, Enterprise, Admin-console, Slackbot. It targets enterprise deployment with authentication and observability features described in the README.
Caveats
- License: Apache-2.0
- Created: 2023-06-20
- Last push: 2024-07-24
- Language: Python
- The repository description mentions finetuning GPT-family models on user data as part of features, per release notes, but no detailed usage guidelines beyond the docker-compose pattern are provided in the excerpt.