Your Data, AI-Accessible

The power of Calliope’s AI tools comes from connecting to your actual data. Not sample data. Not uploaded files. Your live databases and systems.

Here’s how to connect.

Supported Data Sources

Relational databases:

• PostgreSQL
• MySQL / MariaDB
• SQLite
• Microsoft SQL Server

Cloud data warehouses:

• Snowflake
• BigQuery
• Amazon Redshift
• Azure Synapse

File storage:

• Amazon S3
• Google Cloud Storage
• Azure Blob Storage
• Local files

Other sources:

• REST APIs
• CSV/Excel files
• Parquet files

Connection Process

1. Gather credentials

• Host/endpoint
• Database name
• Username/password (or service account)
• SSL requirements

2. Create connection in Calliope

• Navigate to data sources
• Select connection type
• Enter credentials
• Test connection

3. Configure schema access

• Select schemas/tables to expose
• Add descriptions for AI understanding
• Set access permissions

4. Start querying

• Natural language queries work immediately
• AI understands your schema

Secure Credential Management

Credentials are handled securely:

Storage:

• Encrypted at rest
• Never stored in plain text
• Not visible to users after entry

Access:

• Role-based permission to use connections
• No direct credential exposure
• Connection used through Calliope

Audit:

• All data access logged
• Queries recorded for compliance

Schema Documentation

Help AI understand your data:

Table descriptions: “customers: Main customer table with demographic info”

Column descriptions: “customer_id: Primary key” “ltv: Lifetime value in USD” “churn_date: Date customer churned (null if active)”

Relationships: Document how tables relate.

Better documentation = better AI queries.

Access Control

Control who can access what:

Connection level: Who can use this database connection?

Schema level: Which schemas are visible?

Table level: Which tables within schemas?

Row level: Filter data by user context (advanced)

Least privilege. Always.

Query Execution

When users query through AI:

1. Natural language question received
2. AI generates SQL based on schema
3. SQL shown to user (optional)
4. Query executes against database
5. Results returned to user
6. Everything logged for audit

Users get answers. You maintain control.

Performance Considerations

Large databases need consideration:

Query limits: Set row limits to prevent massive queries

Timeout settings: Don’t let queries run forever

Read replicas: Consider pointing at read replicas

Query review: Complex queries may need DBA review

Troubleshooting Connections

Connection refused:

• Check host/port
• Verify firewall/security groups
• Confirm credentials

Permission denied:

• Verify user has SELECT access
• Check schema permissions
• Confirm role assignment

SSL errors:

• Enable/disable SSL as needed
• Check certificate requirements

Timeout:

• Network path open?
• Database responsive?

The Data Connection Checklist

Connecting data sources:

• Credentials gathered (securely)
• Connection created in Calliope
• Connection tested successfully
• Schema documented for AI
• Access permissions configured
• Test query works

Your data, AI-enabled.

Connect your data →