Skip to content

database.md

Latest commit

 

History

History
98 lines (73 loc) · 2.83 KB

database.md

File metadata and controls

98 lines (73 loc) · 2.83 KB
Raw
description Query your data with SQL or natural language.

Database

The Database tab provides direct access to your workspace data through SQL and natural language queries.

Sub-tabs

{% tabs %} {% tab title="SQL Explorer" %} Write and execute SQL queries directly:

  • Full SQL editor with syntax highlighting
  • Query results displayed as a table
  • Export results as CSV
  • Save queries for reuse

{% endtab %}

{% tab title="Semantic Layer" %} Natural language to SQL mappings. Define how business terms map to database columns and tables, so agents can answer data questions without writing SQL.

Example: "How many agents are active?" maps to SELECT COUNT(*) FROM agents WHERE status = 'active' {% endtab %}

{% tab title="Query Templates" %} Pre-built query templates for common operations:

  • Agent performance summaries
  • Document processing statistics
  • Workflow execution history
  • Usage analytics

Click a template to run it instantly or customise it first. {% endtab %}

{% tab title="Training" %} Teach the NL2SQL system with examples:

  1. Enter a natural language question
  2. Provide the correct SQL query
  3. The system learns to handle similar questions

More training examples = better accuracy for your agents' data queries. {% endtab %}

{% tab title="Schema Browser" %} Explore your database schema:

  • Tables and their columns
  • Column types and constraints
  • Relationships between tables
  • Indexes {% endtab %}

{% tab title="Audit History" %} Log of all database queries executed:

  • Query text
  • Who ran it (user or agent)
  • Execution time
  • Result row count {% endtab %} {% endtabs %}

NL2SQL benchmarking

Track how well the natural language to SQL system performs:

  1. Add Golden SQL examples — verified correct query/answer pairs
  2. Run benchmarks against the golden set to measure accuracy
  3. View accuracy trends over time to see if training is improving results

Go to Knowledge Bases → Database → Training to add training examples and golden SQL pairs.

Query templates

The platform ships with 20+ pre-built query templates per database type. These cover common operations like:

  • Active agent counts and performance summaries
  • Document processing statistics
  • Workflow execution history
  • Token usage and cost analytics

Click a template to run it, or customise the SQL before executing.

Query audit

Every database query (whether from a user or agent) is logged with:

  • The natural language question (if applicable)
  • The generated SQL
  • Execution time and row count
  • Who ran it

This audit trail is available in the Audit History sub-tab.

{% hint style="warning" %} SQL queries execute with read-only access by default. Write operations require explicit permission in Settings. {% endhint %}