TattvaDocs
Getting started

Connect data sources

Upload files, see what is connected, and know when a source is ready to answer questions.

Tattva answers questions from sources. A source is a file you upload, a BigQuery table your data analyst connects, or a source shared from the Workspace Library. Every figure in an answer comes from a source.

The Sources page

Sidebar → Sources. Four tabs:

  • All: every source in the project.
  • Upload: add files, and open a source to see its details.
  • Apps: has no connectors to set up. Add data with Upload or BigQuery.
  • BigQuery: tables from your company's warehouse. Usually set up by a data analyst.

All sources

Upload a file

  1. Open the Upload tab.
  2. Click Upload more and pick one or more files. The box accepts XLSX, XLS, CSV, TSV, JSON, Parquet, MD and TXT files.
  3. Each file shows its upload progress, then Queueing…, then Done.

Upload tab

Files you added in the project wizard arrive here the same way.

Metric documents go in the same box

Data analysts upload written metric rules as .md or .txt files through this same box. Tattva turns them into definitions for review. See Write a metric document.

Wait for Ready

After upload, each source runs through four steps, shown on the source:

  1. Parsing: Tattva reads the rows and columns.
  2. Analyzing: it works out what each column means and matches columns to business concepts.
  3. Indexing: it makes the source available to questions.
  4. Ready: you can ask about it.

A small file takes about a minute. If a step fails, you see Pipeline failed — check logs. Try uploading the file again. If it fails again, ask your data analyst.

Connect BigQuery

If your company's data lives in BigQuery, a data analyst connects it on the BigQuery tab. The workspace connects through its own service identity, so there are no credentials to enter. Pick a dataset, pick one or more tables, and click Connect as source. Tattva reads the table's structure. Queries run in BigQuery when you ask a question; the rows are not copied in.

BigQuery tab

See Connect and describe data for the data analyst's view.

Look inside a source

Open a source to see four tabs:

  • Overview: a summary, key findings, the Inspector (granularity, key columns, when it was last evaluated) and Gotchas, which are data issues worth knowing, such as a column with many blanks.
  • Schema: every column with its type, role and meaning.
  • Lineage: what the ontology learned from this source.
  • Queries: example questions this source can answer.

Source overview

Workspace sources

Some sources come from the Workspace Library and carry a Workspace chip. Your project sees them automatically, and your data analysts maintain them. If one is not relevant to your project, hide it with the eye icon on its row in the Upload tab. Other projects keep seeing it. Hidden sources lists what you have hidden; click Show to bring one back.

Tips

  • Upload raw rows, not summaries. One row per order beats a pivot table.
  • Use clear file names. 2026_q1_orders.csv beats data.csv.
  • Read the gotchas. They often explain an odd number before you ask about it.
  • Upload a file with the same name to refresh it. Tattva keeps it as a new version of the same source.

Next: Ask your first question.

Connect data sources — Tattva Docs