AskDB

A database query assistant powered by LangChain and OpenAI. You pass human readable question, it transform it and run, then returns formatted example using LLM agents. You need only to set PostgresSQL uri string in .env and OpenAI key. No schema description is required. Example of the result:

Question: Get me 3 users with the payments with most amount of payments. Give me only name and email.

Here’s a summary of the data which consists of a list of individuals, including their first names and email addresses.
### Summary of Contact Information

| First Name | Email                        |
|------------|------------------------------|
| Charlie    | charlie.davis@example.com    |
| Jane       | jane.smith@example.com       |
| Bob        | bob.wilson@example.com       |

- **Total Contacts:** 3
- **Details:**
  - **Charlie Davis** can be reached at **charlie.davis@example.com**.
  - **Jane Smith** can be reached at **jane.smith@example.com**.
  - **Bob Wilson** can be reached at **bob.wilson@example.com**

Setup

Prerequisites

• Node.js
• Docker and Docker Compose
• OpenAI API key

Installation

1. Install dependencies:

npm install

2. Copy environment file:

cp env.example .env

3. Add your OpenAI API key to .env:

OPENAI_API_KEY=your_api_key_here
POSTGRES_URL=postgresql://postgres:postgres@localhost:5432/askdb

Database Setup

1. Start PostgreSQL using Docker Compose:

npm run docker:up
# or
docker-compose up -d

This will:

• Start a PostgreSQL 15 container
• Create the askdb database
• Automatically run seed.sql to create tables and sample data

2. Test the database connection:

npm run test:db
# or
./test-db.sh

Test Database Schema

It has a seed.sql to make test data for demonstrating the tool. The database includes three related tables:

• users: User information (id, username, email, first_name, last_name)
• user_balances: One-to-one relationship with users (user_id, balance, currency)
• payments: Many-to-one relationship with users (user_id, amount, payment_type, status)

Relationships:

• user_balances.user_id → users.id (one-to-one)
• payments.user_id → users.id (many-to-one)

Docker Commands

# Start database
npm run docker:up

# Stop database
npm run docker:down

## Usage

Run the application with a question as a command line argument:

```bash
npm run dev "Your question here"

For example:

npm run dev "Get me 3 users with the payments with most amount of payments. Give me only name and email."

The application will:

1. Process your question through the SQL execution agent
2. Generate and execute the appropriate SQL query
3. Summarize the results using the summary agent
4. Display the final answer

Note: The question parameter is required. If you don't provide a question, the application will exit with an error message.

Project Structure

src/
  db/              # Database providers
  llm/
    agents/        # LangChain agents (executeSql, summary, pipeline)
    tools/         # LangChain tools (getSchema, makeQuery)
  types/           # TypeScript type definitions