Skip to content

API Reference

This section contains the complete API reference for the LouieAI Python client library.

Overview

LouieAI provides flexible APIs for interacting with the Louie.ai service:

from louieai.notebook import lui

# Ask questions naturally
lui("Show me patterns in my data")
print(lui.text)
df = lui.df

2. Factory Function (New in v0.2.0)

from louieai import louie

# Create callable interface with various auth methods
lui = louie()  # Uses environment variables
lui = louie(username="user", password="pass")  # Direct auth
lui = louie(graphistry_client)  # From PyGraphistry

# Use it naturally
lui("Analyze fraud patterns")

3. Traditional Client API

import louieai
import graphistry

# Authenticate with Graphistry
graphistry.register(api=3, username="your_user", password="your_pass")

# Create client and ask questions
client = louieai.LouieClient()
response = client.add_cell("", "Show me patterns in my data")

# Or use the callable interface (v0.2.0+)
response = client("Show me patterns in my data")

Main Components

Notebook API

Streamlined interface optimized for Jupyter notebooks with implicit thread management and easy data access.

louie() Factory Function

Flexible factory function for creating callable Louie interfaces with various authentication methods.

LouieClient

The traditional client class for interacting with Louie.ai. Provides full control over threads, authentication, and responses. Now also callable!

Available Agents

LouieAI provides 40+ specialized agents with semantic understanding: - General AI: LouieAgent (default), TextAgent, CodeAgent - Databases with Semantic Layer: - PostgresAgent, MySQLAgent, DatabricksAgent, SnowflakeAgent, and more - Agents learn your schema and build semantic models for intelligent query generation - Natural language queries leverage understanding of your business context - Visualizations: GraphAgent, KeplerAgent, PerspectiveAgent, MermaidAgent - Direct Execution: PassthroughAgent variants for each database

# Default agent
lui("Analyze my data")

# Specify a specialized agent
lui("Show user activity", agent="PostgresAgent")
lui("Create network graph", agent="GraphAgent")

Installation

Using uv (recommended):

uv pip install louieai

Using pip:

pip install louieai

Requirements

  • Python 3.10 or higher
  • Active Graphistry account with API access
  • Network access to Louie.ai service

Authentication

Both APIs support multiple authentication methods:

  1. Environment Variables (recommended for notebooks): bash export GRAPHISTRY_USERNAME=your_username export GRAPHISTRY_PASSWORD=your_password

  2. Graphistry Registration: python import graphistry graphistry.register(api=3, username="user", password="pass")

  3. Direct Credentials: python client = louieai.LouieClient(username="user", password="pass")

Error Handling

  • Notebook API: Returns None or empty collections instead of raising exceptions
  • Client API: Raises RuntimeError exceptions on failure

See the respective documentation for detailed error handling examples: - Notebook API Error Handling - Client API Error Handling