Pipeline Quick Start Guide

Overview

Follow these steps to create a project, process a dataset, and download the result.

Before You Begin

Sign in to Rhombus AI before you create a project:

• Go to www.rhombusai.com and click Sign In.
• If you don’t have an account, click Sign Up to create one.

After signing in, Rhombus opens the Workflow page.

Step 1 — Create Your First Project

1. Open the Projects panel on the left.
2. Click New Project.

Name the Project

1. Enter a project name.
2. Click Create.
3. The project appears in the Project List in the left panel.

tip

1. A project contains its datasets, transformations, and results.
2. Free and Starter plans support up to 3 projects, Personal supports up to 10 projects, and Pro and Enterprise support unlimited projects. See current plan details.

Step 2 — Upload Your Dataset

1. In the Chatbot panel, click the “+” button.
2. Select Add new file to upload from your computer. To connect to Amazon S3 (AWS), Azure Blob Storage, Google Cloud Storage, or Snowflake, choose Third party sources.
3. After the upload or import finishes, the file appears in the Chatbot panel, ready for analysis or use in a pipeline.

Third-party source availability

The Third party sources option is available only when the fivetran-3rd-party-data-integration feature is enabled for your account or deployment. Your project must also have an available data-connector slot under its subscription plan.

tip

1. Local file uploads accept CSV, XLS (Excel), and XLSX (Excel).
2. Free plan users can upload files up to 20 MB per file. Upload limits can differ by plan and upload surface.

To use a PDF / Image or Web URL, select a Data Input node and choose the matching option in Upload Data.

Data Input sampling

Data Input nodes always use sampling and memory-safe streaming. By default, the Head method takes the first 100,000 rows. You can change the row count or percentage, or select Head, Systematic, or Reservoir sampling.

Step 3 — Work with the AI Agent

Rhombus uses Analysis mode by default. A regular chat request can analyze your data and create workspace artifacts, but it does not create or change a canvas pipeline. To build one, select Pipeline explicitly.

Step 3A — Analyze and Preview Your Dataset (Default)

1. After uploading the dataset, enter an analysis request in the Chatbot panel. For example: “Analyze this dataset and identify missing values and unusual patterns.”
2. Send the request as usual. To show the /analysis command explicitly, type / and select Analysis.
3. When the agent finishes, its files, tables, and charts appear under Workspace.

info

/analysis is the default, so you do not need to type it for regular analysis requests.

Preview the Analysis Output

1. Open Workspace.
2. Select a file, table, or chart.
3. Check the preview against your request.

tip

1. A dataset preview may show only a subset of the rows.
2. Analysis artifacts are separate from the canvas pipeline. Creating an artifact does not add transformation nodes.

Step 3B — Create a Pipeline

1. Start a new message in the Chatbot panel.
2. Type / and select Pipeline. The Pipeline command label appears in the composer.
3. Describe the pipeline you want. For example: “Create a pipeline that imputes missing values and then removes duplicate rows.”
4. Send the message. The agent creates or updates the canvas nodes.
5. Review the nodes, connections, and configuration.

caution

A request without /pipeline stays in the default Analysis mode and does not create or modify a canvas pipeline.

See Chat Commands for /analysis, /pipeline, /agent, and /plan examples.

Run and Preview the Pipeline

1. In the canvas toolbar, click the play icon beside the + icon. Its hover label is Run Pipeline.
2. Wait for the run to finish.
3. Select the node whose output you want to inspect. A Preview option appears beside Workspace.
4. Click Preview.
5. Check the processed output and adjust the pipeline if needed.

note

/pipeline does not run a new or changed pipeline automatically. Click the play icon to run it and update the preview.

Step 4 — Refine Your Pipeline (Optional)

If the preview needs work, ask the agent to update the pipeline or edit it manually.

Option A — Use the AI Agent

1. Continue the conversation in the Chatbot.
2. Select /pipeline and provide any extra context or instructions.
3. Review the changes.

Option B — Manually Adjust the Pipeline

1. Double-click a transformation node to change its parameters.
2. Add more transformation nodes from the Palette if needed.

Run and Check Your Changes

After any change:

1. Click the play icon beside the + icon.
2. Wait for the run to finish.
3. Select the node whose output you want to inspect.
4. Click Preview beside Workspace.
5. Repeat Step 4 until the output is correct.

Additional Dataset Inspection (Optional)

• Open Dataset Profile in the side panel.
• Review the schema, field types, and statistics.

tip

1. If a node reports an error, click “Fix in Chat” to ask the agent for help.

Step 5 — Download Your Results

Choose the download steps for the type of result you created.

Step 5A — Download a Pipeline Output

1. Run the latest pipeline changes with the play icon.
2. Select the final output node. A Preview option appears beside Workspace.
3. Click Preview.
4. In the Preview panel, open the Download menu.
5. Select Download as CSV or Download as Excel (XLSX).

tip

1. The download contains the final processed dataset produced by the configured transformations.
2. If the result is not what you expected, return to Step 4, update the pipeline, run it again, and review the new preview.

Step 5B — Download an Analysis Artifact

Results from default Analysis mode or an explicit /analysis request are stored in Workspace.

1. Open Workspace to see the artifacts generated by the agent.
2. Select the file you want to inspect.
3. Review its preview.
4. Click Download to save the selected file.