# First tutorial

An optional fictional example of reviewed body versus heading-context retrieval.

CLI version: 0.2.1 · [Read online](<https://ragvidence.medo-id.com/docs/first-tutorial>)

## Run it

**Terminal**

```
ragvi tutorial
```

Creates workspaces/first-run and writes workspaces/first-run/reports/tutorial.html. The command refuses to overwrite an existing workspace. Use ragvi tutorial --name first-run-2 to repeat.

## Expected result

The bundled fictional English exercise compares BM25 body and authentic heading-context representations at k=2. It teaches mechanics; it is not a benchmark for your corpus.

| Run | Sufficient | Rate |
| --- | --- | --- |
| bm25-body | 6 / 10 | 60% |
| bm25-heading\_context | 9 / 10 | 90% |

> **Why the report lists 12 questions**
>
> Ten questions are answerable; two unanswerable questions have N/A retrieval metrics and remain outside the denominator. Three failures recover, none regress, six stay sufficient and one stays insufficient.

## Inspect the evidence

- Check the fixed k and representation.
- Open changed question details and compare ranked source units.
- Check all required evidence, including conditions and exceptions.
- Read the limitation beside the diagnosis.

- [Generated tutorial report from PyPI 0.2.1](<https://ragvidence.medo-id.com/tutorial-report.html>)