Automatic Description Generator
AnalytAI is designed to understand your data with minimal effort from you. The Automatic Description Generator is the engine that scans your raw database schema and translates it into meaningful business descriptions.
How it Works
Section titled “How it Works”When you first connect a data source or upload a file, the generator begins a multi-step process:
- Schema Scan: It reads all table names, column names, and data types from your database.
- Sample Analysis: The AI takes a small, anonymized sample of data from each table to understand what is actually stored there.
- Semantic Mapping: It uses Large Language Models to “guess” the meaning of columns based on their names and samples (e.g., recognizing that
cust_lvllikely means “Customer Subscription Tier”). - Vectorization: The generated descriptions are stored in a specialized “Vector Database” so the AI can quickly find them during your chats.
Why is it Important?
Section titled “Why is it Important?”- Zero Configuration: You don’t have to manually label every column in your 100-table database.
- Contextual Chat: Because the AI knows what a column means, it can answer questions like “Show my most loyal customers” instead of requiring you to say “Filter by cust_lvl > 5”.
- Knowledge Base: The generator creates a searchable library of your company’s data definitions.
Monitoring Progress
Section titled “Monitoring Progress”You can watch the generator at work in your Project Settings → Metadata page. You will see progress bars for:
Generating Table DescriptionsGenerating Column DescriptionsVectorizing Results
Refining the AI’s Understanding
Section titled “Refining the AI’s Understanding”While the generator is highly accurate, you can always improve it:
- Manual Edits: You can manually override any description if the AI missed a nuanced business rule.
- Regenerate: If you significantly change your database structure, you can trigger a “Full Refresh” to update all descriptions.
- Project Context: Use the Project Context page to provide high-level technical notes that help the generator be more precise.
Speed Tip
The initial generation for very large databases (100+ tables) can take a few minutes. Once complete, however, the AI will respond to your chat questions almost instantly!