Component Search
Component Search is the Components view in the dashboard. It searches every component source available to you at once — the standard library, components published to your backend, external libraries you have registered, and your own user components — and presents the matches in a single ranked list.
The same search is available inside the pipeline editor as the Component Search window, so you can find and place a component without leaving the canvas — see The editor window.
Searching
Type into the search box at the top of the view. The placeholder suggests the range of queries that work:
e.g.
train_test_split,pandas,clean up my data
Search matches against several parts of a component, not just its name:
- Component name
- Description
- Inputs and outputs, including their types
- Implementation details
- Metadata
Each result explains why it matched, so a hit on an input type reads differently from a hit on the name. Your query is reflected in the URL, so you can bookmark or share a search.
When nothing matches, the view suggests alternative searches based on the component sources you have loaded.
Sources
Every result carries a badge naming where the component came from:
| Source | Meaning |
|---|---|
| Standard | The curated standard library that ships with TangleML |
| Published | Components published to the connected backend. See Published Components Library |
| User | Your own components |
| Library name | A registered external library, badged with that library's name |
Components that appear in more than one source are shown once, under the highest-priority source.
You can switch individual sources off to narrow a search, and re-enable them all in one action. The active source filters are stored in the URL alongside the query.
AI search
Next to the search box is an AI search button (a sparkle icon). It takes the top candidates from your current search and reranks them with a language model, which helps when you know what you want the component to do but not what it is called.
Reranked results gain a relevance percentage so you can see how confident the ranking is.
AI search requires an OpenAI-compatible provider. Until one is configured, the button is disabled and the view explains that AI search unavailable — ordinary search results are unaffected. Configure a provider under Settings → AI Configuration, where you set the API base URL, API key, and model.
AI search reranks a bounded set of candidates that your query already matched locally. It reorders and explains results — it does not reach beyond the component sources you have loaded.
Component details
Selecting a result opens a detail pane showing the component's description, inputs and outputs, and full source definition. From the pane header you can:
- Copy component link — a shareable URL that reopens the view with this component selected
- Copy component to clipboard — the component definition itself
- Close component details
The pane also shows the component's prefilled description — the description its author wrote — and offers a Generate with AI button that writes an AI description for components whose own documentation is thin. Generating a description is optional and requires the same AI provider configuration as AI search.
The editor window
Inside the pipeline editor, the same search runs in the Component Search window. It is narrower than the dashboard view and geared towards placing components rather than reading about them, so it differs in a few ways.
Browsing
Leave the search box empty and the window shows a browsable tree instead of a result list:
| Folder | Contents |
|---|---|
| Canvas Tools | Sticky Note — an annotation node that carries no execution meaning |
| Inputs & Outputs | Input and Output nodes for pipeline-level parameters and results |
| Your component sources | One folder per source, expandable to the components inside |
Drag anything from the tree onto the canvas to place it. Dropping a component onto an existing task swaps that task's component instead — see Replacing a task's component.
Searching from the editor
Type into the box and the tree is replaced by ranked results, which match against the same fields as the dashboard view. Results appear as you type, without submitting.
Above the results:
- Add Component — opens the Add Component dialog, which adds a component to your user library from a File, a URL, or a New component you write in place. See Adding Components.
- A sparkle button reranks the current results with AI, exactly as AI search does in the dashboard. While it works, the window reports its progress; the button is disabled when no AI provider is configured or when there is nothing to rerank.
- A row of source icons filters results by source. Click an icon to switch that source off, and re-enable them all in one action.
Related components
Alongside search results, the view surfaces two kinds of suggestion to help you assemble a pipeline rather than just find one component:
- Collections — registered libraries that contain matching components, with a count and a preview of the component names inside.
- Compatible components — components whose types line up with your current selection. Each suggestion is labelled Can use outputs (it accepts what your component produces) or Can provide inputs (it produces what your component needs), and lists the matching types.