XLSX & CSV: Two Paths

When you upload a spreadsheet, Thallus offers two ways to use it — as a document for text search, or as a database table for SQL queries. The right choice depends on what you want to do with the data.

The two paths

Document Path Database Path
Document path Database path
What happens Chunked, embedded, searchable as text Loaded as a queryable SQL table
Best for Reports, narrative data, text-heavy spreadsheets Structured data, analytics, filtering
Search method Document search SQL query
Query style "What does the report say about..." "Show rows where revenue > 100k"
Editable No Configurable per table

Document path

Upload through Documents and your spreadsheet is processed like any other file — chunked into sections, embedded for semantic search, and a structured synopsis is generated.

This path is best when your spreadsheet contains descriptive text, commentary, or narrative data where fuzzy, meaning-based search is more useful than exact row-level queries.

For details on how document uploads work, see Uploading Documents.

Database path

Upload through Data Connections as a table and Thallus creates a queryable database table from your file:

Upload file
→
Parse columns
→
Create table
→
Introspect schema
→
Ready to query
  • Column types are automatically inferred — numbers, dates, and text are detected from your data
  • Table name is derived from the filename
  • Schema introspection runs just like it does for external database connections — agents can see columns, types, and row counts
  • Full SQL querying is available — filter, sort, aggregate, and join with other tables

Choosing the right path

A quick guide based on what you're trying to do:

Scenario Best path
"I have a sales report with commentary" Document
"I have 10,000 rows of transaction data" Database
"I want to filter, sort, and aggregate" Database
"I want agents to cite specific paragraphs" Document
"I want to run calculations across rows" Database
"I have a mix of narrative and data" Upload twice — one copy each way

Visibility

Uploaded tables have two visibility settings:

  • Private (default) — Only you can see and query the table
  • Organization-wide — All org members can query the table

You can change visibility at any time from the Data Connections page by selecting the table and updating its settings.