Skip to content

Query Patterns

This guide covers common query patterns for effective use of LouieAI.

Basic Query Structure

LouieAI accepts natural language queries that can include: - Direct questions about your data - Analysis requests - Visualization instructions - Multi-step investigation workflows

Query Types

Data Exploration

  • "Show me the schema of the users table"
  • "What data sources are available?"
  • "Summarize the customer_transactions dataset"

Analysis Queries

  • "Find anomalies in the transaction data from last week"
  • "Calculate the conversion rate by marketing channel"
  • "Identify the top 10 customers by revenue"

Visualization Requests

  • "Create a graph showing user connections"
  • "Plot sales trends over the last 6 months"
  • "Show me a heatmap of activity by hour and day of week"

Investigation Workflows

  • "Investigate suspicious login patterns for user X"
  • "Trace the flow of funds from account A to account B"
  • "Find all related entities connected to IP address Y"

Best Practices

  1. Be Specific: Include relevant details like time ranges, entity names, or specific metrics
  2. Provide Context: Mention the goal of your analysis to get more targeted results
  3. Iterate: Start with broad queries and refine based on initial results
  4. Use Domain Language: Use terminology specific to your data and business domain

Advanced Patterns

Combining Multiple Operations

lui("Find high-value transactions from last month, then visualize the network of entities involved, and identify any suspicious patterns")

Conditional Analysis

lui("If there are more than 100 failed login attempts today, show me the geographic distribution and identify potential attack patterns")

Comparative Analysis

lui("Compare this week's sales performance to the same week last year, highlighting significant changes")

Working with Results

Access previous results using the cursor history:

# Get dataframe from previous query
df = lui[-1].df

# Access multiple past results
for response in lui.history[-5:]:
    print(response.text)

DataFrame Analysis Patterns

Uploading Your Data

Analyze your own pandas DataFrames using natural language:

import pandas as pd

# Load your data
df = pd.read_csv("data.csv")

# Basic analysis
lui("What insights can you find?", df)

# Specific questions
lui("Which columns have missing values?", df)
lui("Show correlation between features", df)
lui("Identify outliers in the sales column", df)

Common DataFrame Operations

# Statistical analysis
lui("Calculate descriptive statistics", df)
lui("Show distribution of values by category", df)

# Data quality checks
lui("Check for data quality issues", df)
lui("Find duplicate rows", df)

# Transformations
lui("Normalize the numeric columns", df)
lui("Create pivot table by region and product", df)

# Time series analysis (if datetime column exists)
lui("Analyze trends over time", df)
lui("Detect seasonality patterns", df)
lui("Forecast next 30 days", df)

Combining Multiple DataFrames

# Analyze first dataset
lui("Profile this customer data", customers_df)

# Add related data for deeper analysis
lui("Now analyze purchase patterns", purchases_df)

# Ask questions across both datasets
lui("Which customer segments have highest lifetime value?")

See Also