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

# Verify in 60 seconds

> Nothing here asks for trust. Every claim is checkable against a live endpoint, a public explorer or a command you run yourself.

## Against the live deployment

<Steps>
  <Step title="Search the catalog">
    ```bash theme={null}
    curl "https://stellarsight.xyz/discovery/search?query=invoice%20ocr&limit=3"
    ```

    Results arrive under `resources`, each with an `_explain` score breakdown —
    the ranking is reproducible, no LLM in the path.
  </Step>

  <Step title="Check the facilitator's contract">
    ```bash theme={null}
    curl "https://stellarsight.xyz/supported"
    ```

    The Stellar extra contract is emitted, including `areFeesSponsored: true` —
    the buyer holds zero XLM.
  </Step>

  <Step title="Read the health of the catalog">
    ```bash theme={null}
    curl "https://stellarsight.xyz/discovery/health"
    ```

    Reports mode, record counts (demo records are labeled `seeded` and never
    count as real), and the exact commit being served.
  </Step>

  <Step title="Open a settled payment">
    Pick any hash in [Settled transactions](/evidence/testnet-transactions) and
    open it on the explorer: `successful: true`, and the `fee_account` is the
    facilitator's fee payer, not the buyer.
  </Step>
</Steps>

## On your machine

```bash theme={null}
git clone https://github.com/pedro-pelicioni/stellarsight && cd stellarsight
npm install && npm run setup
```

No captcha, no faucet, no API key — `setup` issues the SXT test asset and funds
every account the demo needs.

| Command                      | What it proves                                                                         |
| ---------------------------- | -------------------------------------------------------------------------------------- |
| `npm test`                   | 151 tests, 0 failing — 66 of them adversarial catalog-integrity cases                  |
| `npm run verify:api`         | 46 conformance checks, the load-bearing ones driven by the stock `withBazaar()` client |
| `npm run eval:search`        | nDCG\@10 0.864 · Recall\@20 0.905 · MRR\@10 0.920 over a 50-query graded set           |
| `npm run verify:conformance` | An unmodified `@x402/fetch` client completes 402 → sign → settle → 200                 |

The same suite runs in public CI on every push and on a nightly schedule, on
Node 22 and 24 against a real Redis, with a merge-blocking regression gate on
search quality.
