Skip to content

Agent Selection Guide

LouieAI automatically selects the appropriate agents based on your query, but understanding the available agents helps you craft more effective queries.

Agent Categories

Database Agents

For querying structured data in various database systems: - Athena: AWS Athena queries on S3 data - BigQuery: Google BigQuery analytics - Snowflake: Snowflake data warehouse queries - PostgreSQL/MySQL: Traditional relational databases - Splunk: Log analysis and security investigations

Visualization Agents

For creating visual representations of your data: - Graph: Network visualizations using Graphistry - Perspective: Interactive data tables and charts - Kepler: Geospatial mapping and analysis - Mermaid: Flowcharts and diagrams

Code & Development Agents

For executing code and creating notebooks: - Code: Python code generation and execution - Notebook: Jupyter notebook creation and modification

Data Processing Agents

For specialized data operations: - TableAI: Advanced table analysis and transformations - Firecrawl: Web scraping and data extraction

Automatic Selection

LouieAI uses several factors to select agents:

  1. Query Keywords: Specific terms trigger relevant agents
  2. "graph" or "network" → Graph agent
  3. "map" or "geographic" → Kepler agent

  4. "code" or "python" → Code agent

  5. Data Context: The type of data influences agent selection

  6. Graph data → Graph visualization
  7. Tabular data → Perspective or TableAI

  8. Log data → Splunk or appropriate database agent

  9. Task Type: The requested operation determines agents

  10. Analysis → Database + Code agents
  11. Visualization → Appropriate viz agent
  12. Investigation → Multiple coordinated agents

Manual Agent Hints

While automatic selection usually works well, you can hint at specific agents:

# Explicitly request graph visualization
lui("Show this data as a network graph")

# Request code generation
lui("Write Python code to analyze this dataset")

# Ask for geographic visualization
lui("Map these IP addresses to show geographic distribution")

Multi-Agent Workflows

Complex queries often use multiple agents:

# This might use: Database → Code → Graph agents
lui("Query user transactions from the database, calculate risk scores, and visualize the high-risk network")

Performance Considerations

  • Database agents: Best for large-scale data queries
  • Code agents: Flexible but may be slower for simple operations
  • Visualization agents: Require data to be loaded first

Troubleshooting

If the wrong agent is selected: 1. Add more specific keywords to your query 2. Break complex queries into steps 3. Specify the desired output format

See Also