Skip to main content

Sources

Your original data—uploaded files—that Shadowfax never modifies.

What are Sources?​

Sources are your raw data imports. Currently, you can upload CSV, Excel, and Parquet files. Sources represent the starting point for all analysis. Shadowfax treats Sources as immutable—they're read-only and never changed by transformations.

Upload Data Interface

Drag and drop files to import data

Why Sources Matter​

Immutability: Your original data stays pristine. All transformations create new Views—never modify Sources.

Flexibility: Combine multiple file uploads in the same Workbook. Database connections coming soon.

Context: Add descriptions and annotations so the AI understands your data better.

Traceability: Always see where your analysis started. Every View traces back to Sources.

How to Import Data​

Uploading Files​

  1. Click "Add Source" or the + icon in your Workbook
  2. Drag and drop files or click to browse
  3. Supported formats: CSV, TSV, Excel (.xlsx), Parquet
  4. Preview your data before confirming
  5. Watch the upload progress bar

File Upload

Drag and drop files for instant upload

Connecting to Databases (Coming Soon)​

We're building support for live database connections. You'll soon be able to:

  1. Connect to databases like Snowflake, BigQuery, and others
  2. Import tables or views directly
  3. Set automatic refresh schedules
  4. Keep your data current without manual re-uploads

More details coming as this feature launches

Adding Context and Metadata​

After importing, help the AI understand your data:

Dataset Description: Let the AI generate an automatic description, or write your own. This helps the AI know what questions to answer.

Column Context: Add notes explaining what columns mean, especially for:

  • Abbreviations or codes (e.g., "Status codes: 1=Active, 2=Suspended")
  • Similar column names (e.g., "ship_date is when we sent it, delivery_date is when customer received it")
  • Business rules (e.g., "discount_pct should always be between 0 and 100")
  • Data quality notes (e.g., "created_date is sometimes NULL for legacy records")

Data Patterns: Specify formats the AI might not recognize automatically (e.g., "dates are in DD/MM/YYYY format")

Adding Context

Add column-level annotations to guide the AI

Common Use Cases​

  • Upload spreadsheets: Drop your Excel reports directly into Shadowfax
  • Combine datasets: Mix multiple uploaded files in one analysis
  • One-time exploration: Upload CSV exports for quick ad-hoc analysis

Tips & Best Practices​

Name your Sources clearly: "orders_2024.csv" is better than "export_v3_final.csv"

Add context immediately: Spend 30 seconds describing your data. It saves hours of back-and-forth with the AI.

Preview before importing: Check that column names and data types look correct.

Use column annotations: If two columns have similar names, explain the difference. The AI will use this context in all future queries.

Start with clean data when possible: While Shadowfax can handle messy data, cleaner Sources lead to faster results.

Understanding Your Data​

After import, use the Profile view to see:

  • Row and column counts
  • Data types per column
  • Summary statistics
  • Null value percentages
  • Sample values

Click the "Profile" toggle at the bottom panel to inspect metadata.

Data Profile View

Profile view shows column statistics and data quality info