Pipeline Map

What the Pipeline Map shows

The Pipeline Map draws every object in your workspace as a node and every configured connection between them as a line. It answers questions that are painful to answer by clicking through detail pages: what feeds this flow, which Cleaner do these two flows share, and which Bucket is nothing writing to.

Open it from Pipeline Map in the sidebar. It is a full page of its own (at /pipeline-map) that fills the screen, with a back arrow to the dashboard. It reads from your live configuration, so it is always current rather than a diagram somebody drew once and stopped updating.

The Tavnit Pipeline Map in Columns layout. An Input column holds two Splitters and two Collections, a Processing column holds one flow and two Cleaners, an Audio column holds a Signal, and a Data and Activity column holds three Buckets, with connecting lines showing how documents move between them. — Columns layout. Documents move left to right: input, processing, then storage.

Nodes and layers

The map shows ten kinds of object, grouped into the same layers as the sidebar and ordered the way documents travel: what brings a document in on the left, what processes and checks it in the middle, and what stores the result on the right.

Layer	Nodes
Input	Splitters and Collections
Processing	Flows, Cleaners and Agents
Intelligence	Matchers, Inspectors and Fillers
Audio	Signals
Data & Activity	Buckets

Archived flows, Splitters, Cleaners, agents, Signals and Buckets are left out. Agents and Signals appear only if they are enabled for your organisation. Subjects, Nets, pipelines and individual runs are not drawn on the map.

How the lines are drawn

Every line comes from a setting you made somewhere else. Hover a node and each of its lines is labelled with the relation it stands for.

Line	Label	Comes from
Collection → Flow	routes to	The flow is in the Collection, or is its fallback flow.
Collection → Splitter	routes to	The Splitter is in the Collection.
Splitter → Flow or Collection	sends documents to	A document type in the Splitter is sent to that flow or Collection.
Flow → Cleaner	cleans with	The flow's data cleaning uses that Cleaner.
Flow → Bucket	exports to	The flow's Bucket Export points at that Bucket.
Flow → Agent	triggers	The flow is linked to that agent.
Flow → Matcher	compared by	The Matcher compares that flow's runs.
Flow → Inspector	checked by	The flow is one of the Inspector's inputs.
Flow → Filler	feeds	The flow is one of the Filler's inputs.
Agent → Bucket	delivers to	The agent has a Bucket delivery.
Signal → Bucket	exports to	The Signal exports to that Bucket.
Cleaner → Bucket	writes to	The Cleaner has a Bucket Check field on that Bucket.

A flow's lines get thicker the more runs it had in the last 7 days, and briefly pulse when a new run starts, so you can see which paths are busy without leaving the map.

Three layouts

The same graph can be drawn three ways. Switch between them from the toolbar or with the keys 1, 2 and 3; none of them changes your configuration, only how it is arranged on screen. The map remembers the last layout you used.

Layout	How it looks	Best for
Columns	The default. One column per node type inside its layer band, each with a count and its own sort. A column shows up to 30 nodes, then a “Show all” control.	Reading the whole wiring left to right, and explaining the setup to someone else.
Lanes	One row per flow or Signal: what feeds it on the left, what it sends to on the right, grouped by relation. Nodes that touch no flow or Signal are listed under “Not connected”.	Following one flow end to end when many objects share the same stage.
Orbits	Each connected flow or Signal is a hub with its direct neighbours on rings around it, inputs on the left and outputs on the right. The most connected hubs come first.	Spotting the busiest hubs, which flow everything else clusters around.

Exploring a node

Hover a node to see its card: whether it is active, its email trigger address, email output, webhook, human review, agent delivery, completion mode (Inspectors and Fillers) or private status (Buckets), plus its number of connections and creation date
Click a node to focus it: its direct connections light up and the card stays pinned with an “Open” button. Click it again, press Esc or click empty space to “Exit focus”
Double-click a node, or press Enter while it is focused, to open its detail page
With a node focused, Shift+click another node to highlight the shortest path between them
Right-click a node for “Open”, “Focus”, “Hide element” (that node type) and “Hide layer”
Use the arrow keys to move between nodes, and / to jump to search

Finding things in a large workspace

Once a workspace has a few dozen objects the whole graph stops fitting on screen. The toolbar has the tools for that.

“Search nodes...” to find one by name. Search hits always show, even past a column's limit
“Filters” opens a panel with Layers and Elements to hide, Configuration filters (Active, Human review, Webhook, Email), and Options: “Connected only” and “Hide non-matching”. Without the last one, non-matching nodes are dimmed rather than removed. “Clear all” resets everything
Sort each column by Connected First, Name A-Z, Name Z-A or Last Created, or sort every column at once
In Columns, zoom in and out (or use the mouse wheel) and click “Fit” to frame the whole graph; a minimap in the corner shows where you are
“Export SVG” downloads the current Columns or Orbits view as an image
Switch between bright and dark mode for the map

Your filters and sort are remembered in your browser. The address bar keeps the layout, search and focused node, so you can share a link that opens the map framed the same way. On small screens the map shows a compact list by layer instead of the canvas.

Finding orphans

Sort by Connected First and read the bottom of each column, or turn on Connected only to see what remains. A flow with no Cleaner and no Bucket is throwing its results away unless something collects them over the API or a webhook; a Bucket with nothing pointing at it is only being written to by hand or over the API.

Pipeline Map and Pipelines

The map shows the connections that already exist in each feature's own settings. A pipeline is different: a canvas you build on purpose to chain steps together and run them as one. Pipelines are not drawn on the map; use the map to understand how your workspace is wired, and Pipelines to orchestrate a specific sequence.

When the map is worth opening

The map is a diagnostic tool rather than something you use daily. It earns its keep when you need to understand the shape of the workspace rather than the contents of one run.

Before deleting anything: the map shows what else is wired to it
When a document ends up in the wrong place and you need to see the routing path it took
When onboarding someone, as a one-screen explanation of how the workspace fits together
After a build-out, to check nothing was left unconnected
When two teams have each configured flows and you want to find duplicated work

Structure, not results

Apart from line thickness, the map shows how things are connected, not what flowed through them. For individual results and failures, use Runs and each object's own history.

Troubleshooting

Problem	What to check
The map says there is nothing to map	Create a flow, Bucket or Cleaner first; the map only draws objects that exist.
An object is missing	Check that it isn't archived, that its layer or type isn't hidden in Filters, and that “Connected only” or “Hide non-matching” isn't removing it. Agents and Signals only show when enabled for your organisation.
A column ends with “+N”	Columns show 30 nodes each. Click “Show all”, or search for the name.
Zoom and Fit are missing	They are only available in the Columns layout.