GluonDB Agent

GluonDB Agent is an AI-powered assistant that enables natural language interactions with your data. It can analyze datasets, generate visualizations, create dashboards, and execute SQL queries through conversational requests.

---

What the Agent Can Do

The GluonDB Agent is designed to help users interact with their data through natural language. The agent can:

Data Analysis

• Execute SQL queries: Run queries against connected datasources (PostgreSQL, DuckDB, MySQL)
• Explore data: Inspect table schemas, column types, and relationships
• Statistical analysis: Generate descriptive statistics, identify patterns and trends
• Data summarization: Create concise summaries of large datasets

Visualizations

• Generate interactive charts: Create bar, line, pie, scatter, area, and other chart types
• Custom styling: Apply dark mode themes with professional color palettes
• Multiple libraries: Support for Chart.js, D3.js, Plotly.js, and ApexCharts

Dashboards

• Multi-component layouts: Combine KPIs, charts, and tables in unified views
• Automatic file generation: Save dashboards as standalone HTML files
• Responsive design: Generate layouts that adapt to different screen sizes

File Operations

• View and edit files: Read, create, and modify project files
• Navigate file structure: Explore project directories
• Generate reports: Create HTML files with analysis results

Memory and Context

• Session continuity: Maintain conversation context within a session
• Long-term memory: Remember user preferences and important facts across sessions
• Query history: Track executed queries for reference and dashboard generation

---

How to Use

Starting a Conversation

Open the chat interface and type your request in natural language. The agent understands context from previous messages, so you can build on prior interactions.

Example requests:

Query Execution

To run SQL queries, simply describe what data you need:

The agent will:

1. Identify the appropriate datasource
2. Generate and execute the SQL query
3. Return the results with relevant insights

For direct SQL execution, you can also specify the query:

Creating Visualizations

Request charts by describing what you want to see:

The agent will:

1. Fetch the required data
2. Generate the appropriate chart configuration
3. Create an interactive HTML visualization
4. Save the file and display it in the chat

Building Dashboards

For comprehensive views, request a dashboard:

The agent will generate a complete HTML dashboard with multiple components arranged in a professional layout.

Reviewing History

You can ask about previous queries in your session:

---

Scope and Capabilities

Supported Datasources

Supported Chart Types

• Line charts
• Bar charts (horizontal and vertical)
• Pie and donut charts
• Scatter plots
• Area charts
• Combo charts
• Gauge charts
• Heatmaps

--

Limitations

While the agent is capable of many tasks, there are some limitations to be aware of:

Data Processing

• Result set limits: Large queries may be truncated to prevent memory issues (default: 1000 rows)
• Complex joins: Very complex multi-table joins may require manual SQL refinement
• Real-time data: The agent queries data at request time; it does not maintain live connections

Visualization

• Custom styling: While the agent follows a design system, highly specific styling requests may need manual adjustment
• Complex interactivity: Advanced interactive features (linked charts, custom tooltips) may have limited support
• Export formats: Charts are generated as HTML; direct PNG/PDF export is not currently supported

Analysis

• Statistical depth: Advanced statistical tests (regression, hypothesis testing) have limited support
• Machine learning: The agent does not perform ML model training or prediction
• Large datasets: Analysis is optimized for datasets that fit in memory

General

• Language: The agent works best with English prompts, though it may understand other languages
• Ambiguous requests: Vague or incomplete requests may require clarification
• Schema knowledge: The agent must query the datasource to understand table structures; it does not have pre-existing schema knowledge

---

Best Practices

To get the best results from GluonDB Agent:

Be Specific

Clear, specific questions yield better results:

Provide Context

Reference specific tables, columns, or previous results when relevant:

Iterate with Follow-ups

Build on previous responses to refine your analysis:

Use Datasource Information

When working with new databases, start by exploring the schema:

Verify Results

For important analysis, verify the underlying queries:

---

Troubleshooting

Common Issues

"Tool timed out"

The query or operation took longer than expected. Try:

• Adding more specific filters to reduce data volume
• Breaking complex operations into smaller steps

"No data returned"

The query executed but returned no results. Verify:

• Table and column names are correct
• Filter conditions match existing data

"Unknown datasource"

The specified datasource was not found. Use:

Empty or incomplete charts

The visualization may have insufficient data. Try:

• Verifying the underlying query returns data
• Specifying explicit column mappings

Getting Help

If you encounter issues:

1. Rephrase your request: Try a more specific or differently worded prompt
2. Check the schema: Ensure you are referencing correct table and column names
3. Break it down: Divide complex requests into simpler steps
4. Ask for clarification: Request the agent explain what went wrong

---

Related Resources

• Getting Started - How to connect your database to the agent
• gluonDB SDK - How to build custom dashboards and visualizations