Reference

How the canvas works

The story canvas is the front door. Each card is one decision. The table map is a second view of the local tables.

Choose folder

The output folder holds pipelines and exports. The local tables being queried stay in this tab.

Drop a file

A dropped file becomes a local table. The left side opens on Nodes, so a card can be dragged onto the canvas.

Profile rail

Clicking a card shows the profile saved the last time that card ran. Selecting the card does not scan the table again.

Grid

Double-click a card to open the grid. Expand on the card shows the saved rows without scanning again.

Table map

A read-only view of files you dropped and tables a merge produced. A line is a join key. Double-click a box to return to that card.

Suggestions

Chips are counted from the table: empty cells, repeated rows, and a column with one value. Accepting a chip adds one line. Review opens the full list.

Persona

A persona is a name, instructions, and reference notes. It does not hold the rows. Choose it on the agent and in the wrangler chat. Build one on the personas page.

Privacy

The agent sees column names and the profile. It recommends a step. The table changes when you accept the step.

Offline

After one online visit, the dashboard shell can open with the network off. The rows stay on this computer.

Online connectors

API and storage cards send or fetch rows. They are marked online.

Document Extractor

Extract tabular data from PDFs, invoices, and unstructured docs

Document Extractor: the cards connected, the folder chosen, and the rows this step kept

Dataset

Upload, preview, and inspect your data

Dataset: the cards connected, the folder chosen, and the rows this step kept

Clean Data

AI-guided cleaning with 35 rule templates

Clean Data: the cards connected, the folder chosen, and the rows this step kept

Imputation

Fill missing values with KNN or fast methods

Imputation: the cards connected, the folder chosen, and the rows this step kept

Derive Variables

Create calculated columns with formula builder

Derive Variables: the cards connected, the folder chosen, and the rows this step kept

Filter

Filter rows by column conditions

Filter: the cards connected, the folder chosen, and the rows this step kept

Entity Resolution

Deduplicate records with privacy-first matching

Entity Resolution: the cards connected, the folder chosen, and the rows this step kept

Merge / Join

Combine multiple datasets by appending or joining

Merge / Join: the cards connected, the folder chosen, and the rows this step kept

Synthetic Data

Generate synthetic records with 4 methods

Synthetic Data: the cards connected, the folder chosen, and the rows this step kept

Export

Download processed dataset

Export: the cards connected, the folder chosen, and the rows this step kept

Validate

Flag rows that fail quality checks

Validate: the cards connected, the folder chosen, and the rows this step kept

API Source

Fetch data from REST APIs

API Source: the cards connected, the folder chosen, and the rows this step kept

API Destination

Push data to REST APIs

API Destination: the cards connected, the folder chosen, and the rows this step kept

Storage Source

Read from S3/GCS/R2 object storage

Storage Source: the cards connected, the folder chosen, and the rows this step kept

Storage Destination

Write to S3/GCS/R2 object storage

Storage Destination: the cards connected, the folder chosen, and the rows this step kept

Integrations