urimai 0.1.0

AI-powered SQL answer engine - ask questions about your databases

SQL Answer Engine

Ask natural-language questions about SQLite databases and CSV files using AI.

Prerequisites

• Python >= 3.11.6
• uv package manager
• A Google API key (Gemini)

Installation

git clone <repo-url>
cd sql_answer_engine
uv sync

Create a .env file in the project root:

GOOGLE_API_KEY=your-google-api-key

Environment Variables

Variable	Required	Default	Description
GOOGLE_API_KEY	Yes	—	Google Gemini API key
MODEL_NAME	No	gemini-2.5-flash	Model to use for AI queries
METADATA_DB_PATH	No	./metadata.db	Path to the internal metadata database
DEFAULT_SAMPLE_ROWS	No	5	Number of sample rows extracted per table during init/sync

CLI Commands

sql-engine init <path>

Register a SQLite database or CSV file.

sql-engine init ./my_database.db
sql-engine init ./data.csv --name my_data --table-name sales --delimiter "," --encoding utf-8

Option	Description
--name	Custom name for the database (default: filename)
--table-name	Table name for CSV import (default: LLM-suggested)
--delimiter	CSV delimiter character (default: ,)
--encoding	CSV file encoding (default: utf-8)

CSV files are automatically converted to SQLite with LLM-powered schema inference.

sql-engine list

List all registered databases.

sql-engine list

sql-engine chat <db_name>

Start an interactive chat session with a registered database.

sql-engine chat my_database

sql-engine sync <db_name>

Re-sync schema metadata for a registered database. Useful after the source data changes.

sql-engine sync my_database

sql-engine config

View or modify settings.

sql-engine config --show
sql-engine config --sample-rows 10

Option	Description
--show	Show current settings
--sample-rows	Set number of sample rows extracted per table

Quick Start

# Register a database
sql-engine init ./chinook.db

# Start chatting
sql-engine chat chinook