Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.tensorstax.com/llms.txt

Use this file to discover all available pages before exploring further.

Semantic Search in Action

​
Overview

Semantic Search provides an intelligent search experience that goes beyond simple file name matching. Search for errors, ask questions about models, or find files by context - all processed in real-time with no pre-built indexes.

​
How It Works

The search bar understands natural language queries and semantic context:
  • Error Search: Type error messages to quickly locate the problematic files
  • Contextual Queries: Ask questions about models and their functionality
  • File Name Search: Traditional file name searches work seamlessly
  • Macro Discovery: Find macros by describing what they do

​
Performance

Under 500ms - Fast results for typical dbt projects
Under 5 seconds - Even projects with 1000s of files return results quickly

​
Real-Time Processing

No indexes stored - Everything is computed in memory to ensure the most up-to-date results. Changes to your project are immediately searchable without rebuilding indexes.

​
Search Examples

​
Error Detection

  • Type error messages from your dbt runs
  • The search will identify which specific files contain the errors
  • Get direct navigation to problematic code

​
Model Questions

  • “Find the model that calculates customer lifetime value”
  • “Show me models that use the orders table”
  • “Which model handles revenue calculations”

​
Macro Discovery

  • “Find macro that formats dates”
  • “Show me the macro for data quality checks”
  • “Macro that handles null values”
  • Standard file name searches work alongside semantic queries
  • Partial names and patterns are supported
  • Case-insensitive matching

​
Key Features

  • Context-Aware: Understands the purpose and functionality of your models
  • Error-Focused: Quickly pinpoint files causing build failures
  • No Maintenance: No indexes to rebuild or maintain
  • Always Current: Searches reflect your latest code changes immediately
  • Natural Language: Ask questions in plain English
Try combining different search approaches - you can search for errors, then follow up with contextual questions about the affected models.