> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bensevern.dev/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Web workbench (no code)

> The point-and-click way to use GoldenMatch: edit rules, preview matches, label pairs, and compare runs in a browser — no code after the first install.

If you would rather not write code, GoldenMatch ships a browser workbench. You (or
a teammate) run one command to start it; after that, everything happens
point-and-click in your browser. It is the right starting point for analysts,
data stewards, and anyone reviewing matches rather than building a pipeline.

## Start it

<Steps>
  <Step title="Install the web extra">
    ```bash theme={null}
    pip install 'goldenmatch[web]'
    ```

    One-time, and the only command line you need. If someone technical set this up
    for you, skip to the browser.
  </Step>

  <Step title="Open the workbench">
    ```bash theme={null}
    goldenmatch serve-ui my-project
    ```

    This opens **[http://localhost:5050](http://localhost:5050)** in your browser. Point it at a folder of
    CSVs or an existing project; it picks a free port automatically if 5050 is taken.
  </Step>

  <Step title="No data yet? Try the demo">
    ```bash theme={null}
    goldenmatch demo
    ```

    Generates a small sample dataset and runs it, so you can click around before
    bringing your own file.
  </Step>
</Steps>

<Tip>
  Drop a CSV in and run it with zero configuration — GoldenMatch detects the column
  types, picks the scorers and blocking, and shows you the matches. You only touch
  the rules if you want to.
</Tip>

## What you can do in the browser

No code required for any of this:

* **See your duplicates.** Browse clusters of records GoldenMatch thinks are the
  same entity, with a field-by-field diff and a plain-English reason for each pair.
* **Tune without YAML.** Edit match rules and thresholds with live validation and
  preview the effect against a sampled slice before committing to a full run.
* **Label pairs.** Mark borderline pairs as match / not-a-match. Your decisions are
  saved to Learning Memory automatically and re-applied on the next run, so the
  system stops asking you the same question twice.
* **Compare two runs.** See what merged, split, or stayed the same between runs
  (the CCMS view), so you can tell whether a change helped.
* **Sweep a setting.** Move the threshold and watch the cluster count update live.
* **Browse corrections.** Review and manage every label you have stored.

## When you outgrow it

The workbench writes the same config and corrections the code paths use, so nothing
is locked in. When you are ready for automation or scale:

<CardGroup cols={2}>
  <Card title="Command line" icon="terminal" href="/docs/goldenmatch/cli">
    Run the same matching from a script or a scheduled job.
  </Card>

  <Card title="Quickstart (developers)" icon="rocket" href="/docs/quickstart">
    The Python / TypeScript / CLI path, install to deduped file in minutes.
  </Card>
</CardGroup>

<Note>
  The workbench is a single-process localhost app — your data never leaves your
  machine. More screenshots and a full tour are on the
  [Web UI wiki page](https://github.com/benseverndev-oss/goldenmatch/wiki/Web-UI).
</Note>