Sources & EvidenceSources are evidence attached to Business Questions. They replace the older "dataset-only" model with a more flexible evidence system that accepts CSVs, text, pasted content, and database connections.Source TypesCSV UploadsCSV files receive special treatment in Learning Nodes:
Automatic schema detection: Column names and types inferred on upload
Row/column counts: Displayed in the source preview
Sample preview: First 10 rows shown for verification
AlaSQL query support: Run SQL queries directly in the browser
Cohort creation: Query results can be saved as named cohorts
Text / Paste SourcesFirst-class evidence for unstructured content:
Emails: Paste important email threads as evidence
Meeting notes: Capture decisions from meetings
Research excerpts: Quotes from articles, reports, or documentation
Manual assertions: Expert judgment with rationale
Text sources show word count and are searchable. They can be attached to multiple questions and referenced in decision rationale.Plain Text FilesUpload .txt files containing:
Log extracts
Configuration dumps
Exported chat transcripts
Any text-based evidence
Database ConnectionsAdmin-configured connections to:
PostgreSQL databases
MySQL databases
Data warehouses (BigQuery, Snowflake, etc.)
Users can reference tables from these connections without seeing connection secrets. Query access is controlled at the connection level.Source-Question RelationshipsSources have a flexible relationship model:
One Source → Many BQs: A source can be attached to multiple questions
One BQ → Many Sources: A question can have multiple sources attached
This enables reuse of evidence across related decisions. For example, a quarterly revenue CSV might support multiple Business Questions about different product lines.Evidence & Sources Section in ASKIn the ASK canvas, the Evidence & Sources section provides:
Drag-and-drop upload: For CSV and TXT files
Paste text area: For quick evidence capture
Link from library: Attach existing sources from your workspace
Source MetadataEvery source tracks:
Name: Human-readable identifier
Type: CSV, text/paste, db_connection, etc.
Content: The actual data (for text/paste) or reference (for DB)
Schema info: For CSVs (column names, types)
Sample/Preview: Representative view of the data
Owner: Who uploaded/created the source
Metadata: File size, word count, row count, upload timestamp
Inspecting Sources in DECIDEThe DECIDE canvas includes an Inspect Data panel for CSV sources:
Schema view: Column names, types, sample values
SQL query console: Write and execute AlaSQL queries
Results table: Query output with sorting
Save as cohort: Persist query results for analysis
Attach as evidence: Link findings to checks
Data Quality FlagsSources may show quality indicators:
Freshness warning: "Last refreshed X days ago — results may be stale"
Completeness: "3 columns have >10% null values"
Health banner: Single-line summary in DECIDE (only shown when issues exist)
Healthy sources show no banner (zero clutter principle).Sources IndexThe Sources index (global navigation) shows all sources in your workspace: