Skip to main content
Pipelines let you orchestrate any tools in Tamarind’s catalog together, funneling the outputs of one tool into the inputs of the next to build complex workflows. This tutorial walks through the UI end-to-end: finding pipelines, building one on the canvas, wiring up settings, then testing and publishing it. For concepts, schema, and API examples, see the Pipelines reference.

​
Getting started

Pipelines live in their own tab in the sidebar, and every pipeline you’ve published to your organization also appears alongside the built-in tools on the All Tools page. The Pipelines dashboard is the jumping-off point for creating a new template, opening a starter, or reviewing test and published runs. Left sidebar of the Tamarind app with the Pipelines tab (1) numbered, next to the All Tools page filtered to the Pipelines chip (2).
  1. Build and access your pipelines from the Pipelines tab
  2. Or just see your published pipelines in card format from the Tools page
The Pipelines dashboard: search bar (1), scope filter (2), Templates vs Runs tabs (3), starter template + New pipeline buttons (4), draft vs published filter (5), and per-card action menu (6).
  1. Search your pipelines
  2. Filter between your pipelines and those created by others in your organization
  3. View your pipeline templates or the test/real runs
  4. Start from a starter template for the most used workflows or build your own
  5. View draft or published pipelines
  6. Additional options on template card menu

​
Building a pipeline

A pipeline is a graph of nodes on a canvas: input nodes at the top feed molecules into tool nodes, which pass their outputs downstream to other tools. You can drag tools in from the sidebar, connect ports by hand, or ask the pipelines agent to build a first draft that you refine from there. Empty pipeline editor: the canvas center (1), the tool search sidebar on the left (2), and the pipelines agent chat panel on the right (3).
  1. Your pipeline gets built on the canvas
  2. You can either search/drag your tool nodes
  3. Or ask the pipelines agent to build your pipeline for you
Canvas with an RFdiffusion → ProteinMPNN → AlphaFold/ESM2 pipeline in progress: the input node at the top (1) and the downstream tool nodes below it (2).
  1. Pipelines start with input node(s). These must be populated with a default molecule (protein, small molecule, or nucleic acid) for testing, but when you submit the pipeline you’ll be able to give other inputs
  2. You will then have individual tool nodes that need to be connected together
A single RFdiffusion node: input port at the top labeled with its type (1), the + menu below the node for adding downstream tools (2), and the node's status bar showing molecule and job counts (3).
  1. You can see the input type
  2. And output type. Note that most tools produce multiple output types. Clicking on the tool and going to the Info tab will show you information about the tool’s purpose/input and output specs. You can also add connected nodes using the (+) menu.
  3. The bottom bar shows, based on your settings, how many input molecules are intaken/output molecules are generated and how these are batched into jobs
Wired-up canvas with a node showing an error indicator (1), a per-node settings-needed banner (2), and the errors panel opened from the top-right (3).
  1. Connect nodes together by dragging between inputs and outputs
  2. See if settings modifications need to be made per node to validate the pipeline
  3. Or look at all the detected errors in the top right corner

​
Specifying settings

Every node has its own settings dialog with two audiences. Behind-the-scenes settings are the values you lock in as the author, and editable settings are the ones you expose to whoever runs the pipeline later. Some tools also need you to pick specific chains from an upstream molecule to feed forward. ProteinMPNN node settings dialog: "needs fixing" banner at the top (1), the editable-user-settings picker (2), and the Model Type selector for behind-the-scenes tool settings (3).
  1. The node interface shows what needs fixing
  2. Select what settings the users can edit for each tool when running the pipeline
  3. Go on to define the behind-the-scenes settings for the tool
ESM2 node with the input Protein Sequence field expanded, showing a chain selector for the upstream ProteinMPNN output (1).
  1. Some tools will require selecting certain chain outputs
Input node configuration modal: molecule-type tabs — Protein / Small molecule / Nucleic acid (1), a "requires structure" toggle (2), and a PDB fetch field for the default molecule (3).
  1. For input nodes, select molecule type
  2. Determine whether downstream nodes require structure or just sequence
  3. Add a default molecule like you would in any other Tamarind tool

​
Running a pipeline

Test runs let you validate the whole graph against your default molecule before you commit. Once it looks right, publish the pipeline so others in your organization can run it against their own inputs, and track progress node-by-node from the run view. Pipeline editor with the Test button highlighted in the top-left (1), opening a test-run submission dialog.
  1. Test run a pipeline to ensure everything works properly with the Test button in the top left
Pipeline in Running state: test / published run tabs on the left (1), and a currently-running node highlighted on the canvas (2).
  1. View your test or published runs, or view/modify the pipeline to create a new version
  2. See how each node is progressing, or if there are runtime issues with any node
Publish settings dialog: the publish toggle (1), a link to the submission page (2), and the version picker (3).
  1. Select the publish button in the top left to publish the pipeline for others in your organization to use
  2. The link will take you to your published pipeline, which also shows up in your Tools page on the pipelines tab as well
  3. You can select which pipeline version to publish
My Results tab filtered to "pipeline" — completed pipeline runs listed alongside other jobs, with type, status, and submission time. Your pipelines runs show up in your Results tab like other jobs.