Metadata-Version: 2.4
Name: sqlsaber
Version: 0.22.0
Summary: SQLsaber - Open-source agentic SQL assistant
License-File: LICENSE
Requires-Python: >=3.12
Requires-Dist: aiomysql>=0.2.0
Requires-Dist: aiosqlite>=0.21.0
Requires-Dist: asyncpg>=0.30.0
Requires-Dist: cyclopts>=3.22.1
Requires-Dist: fastmcp>=2.9.0
Requires-Dist: httpx>=0.28.1
Requires-Dist: keyring>=25.6.0
Requires-Dist: pandas>=2.0.0
Requires-Dist: platformdirs>=4.0.0
Requires-Dist: prompt-toolkit>3.0.51
Requires-Dist: pydantic-ai
Requires-Dist: questionary>=2.1.0
Requires-Dist: rich>=13.7.0
Description-Content-Type: text/markdown

# SQLsaber

> SQLsaber is an open-source agentic SQL assistant. Think Claude Code but for SQL.

![demo](./sqlsaber.gif)

Stop fighting your database.

Ask your questions in natural language and `sqlsaber` will gather the right context automatically and answer your query by writing SQL and analyzing the results.

## Table of Contents

- [Features](#features)
- [Installation](#installation)
- [Configuration](#configuration)
  - [Database Connection](#database-connection)
  - [AI Model Configuration](#ai-model-configuration)
  - [Memory Management](#memory-management)
- [Usage](#usage)
  - [Interactive Mode](#interactive-mode)
  - [Single Query](#single-query)
  - [Resume Past Conversation](#resume-past-conversation)
  - [Database Selection](#database-selection)
- [Examples](#examples)
- [MCP Server Integration](#mcp-server-integration)
  - [Starting the MCP Server](#starting-the-mcp-server)
  - [Configuring MCP Clients](#configuring-mcp-clients)
  - [Available MCP Tools](#available-mcp-tools)
- [How It Works](#how-it-works)
- [Contributing](#contributing)
- [License](#license)

## Features

- 🔍 Automatic database schema introspection
- 🛡️ Safe query execution (read-only by default)
- 🧠 Memory management
- 💬 Interactive REPL mode
- 🧵 Conversation threads (store, display, and resume conversations)
- 🗄️ Support for PostgreSQL, SQLite, and MySQL
- 🔌 MCP (Model Context Protocol) server support
- 🎨 Beautiful formatted output

## Installation

### `uv`

```bash
uv tool install sqlsaber
```

### `pipx`

```bash
pipx install sqlsaber
```

### `brew`

```bash
brew install uv
uv tool install sqlsaber
```

## Configuration

### Database Connection

Set your database connection URL:

```bash
saber db add DB_NAME
```

This will ask you some questions about your database connection

### AI Model Configuration

SQLSaber uses Sonnet-4 by default. You can change it using:

```bash
saber models set

# for more model settings run:
saber models --help
```

### Memory Management

You can add specific context about your database to the model using the memory feature. This is similar to how you add memory/context in Claude Code.

```bash
saber memory add 'always convert dates to string for easier formating'
```

View all memories

```bash
saber memory list
```

> You can also add memories in an interactive query session by starting with the `#` sign

## Usage

### Interactive Mode

Start an interactive session:

```bash
saber
```

> You can also add memories in an interactive session by starting your message with the `#` sign

### Single Query

Execute a single natural language query:

```bash
saber "show me all users created this month"
```

You can also pipe queries from stdin:

```bash
echo "show me all users created this month" | saber
cat query.txt | saber
```

### Resume Past Conversation

Continue a previous conversation thread:

```bash
saber threads resume THREAD_ID
```

### Database Selection

Use a specific database connection:

```bash
# Interactive mode with specific database
saber -d mydb

# Single query with specific database
saber -d mydb "count all orders"

# You can also pass a connection string
saber -d "postgresql://user:password@localhost:5432/mydb" "count all orders"
```

## Examples

```bash
# Start interactive mode
saber

# Non-interactive mode
saber "show me orders with customer details for this week"

saber "which products had the highest sales growth last quarter?"
```

## MCP Server Integration

SQLSaber includes an MCP (Model Context Protocol) server that allows AI agents like Claude Code to directly leverage tools available in SQLSaber.

### Starting the MCP Server

Run the MCP server using uvx:

```bash
uvx --from sqlsaber saber-mcp
```

### Configuring MCP Clients

#### Claude Code

Add SQLSaber as an MCP server in Claude Code:

```bash
claude mcp add sqlsaber -- uvx --from sqlsaber saber-mcp
```

#### Other MCP Clients

For other MCP clients, configure them to run the command: `uvx --from sqlsaber saber-mcp`

### Available MCP Tools

Once connected, the MCP client will have access to these tools:

- `get_databases()` - Lists all configured databases
- `list_tables(database)` - Get all tables in a database with row counts
- `introspect_schema(database, table_pattern?)` - Get detailed schema information
- `execute_sql(database, query, limit?)` - Execute SQL queries (read-only)

The MCP server uses your existing SQLSaber database configurations, so make sure to set up your databases using `saber db add` first.

## How It Works

SQLsaber uses a multi-step agentic process to gather the right context and execute SQL queries to answer your questions:

![](./sqlsaber.svg)

### 🔍 Discovery Phase

1. **List Tables Tool**: Quickly discovers available tables with row counts
2. **Pattern Matching**: Identifies relevant tables based on your query

### 📋 Schema Analysis

3. **Smart Schema Introspection**: Analyzes only the specific table structures needed for your query

### ⚡ Execution Phase

4. **SQL Generation**: Creates optimized SQL queries based on natural language input
5. **Safe Execution**: Runs read-only queries with built-in protections against destructive operations
6. **Result Formatting**: Presents results with explanations in tables

## Contributing

If you like the project, starring the repo is a great way to show your support!

Other contributions are welcome! Please feel free to open an issue to discuss your ideas or report bugs.

## License

This project is licensed under Apache-2.0 License - see the LICENSE file for details.
