# Inspect evidence

> Open the recorded source behind a citation, then save and cite your own PDF and DOCX files.

Status: Available

## The recorded source

The staffing objection cites line 4 of the fictional `context.md`. Open that saved revision:

```sh doc-test
node dist/cli.js evidence --line 4 --data-dir .boardroom/example
```

The output shows the saved revision, its location and its SHA-256.

## Save a PDF or DOCX file

Authorize the single file you want BOARDROOM to read with `--allow-source`:

```sh doc-test
node dist/cli.js document --source assets/validation/launch.pdf --allow-source --data-dir .boardroom/documents
node dist/cli.js document --source assets/validation/launch.docx --allow-source --data-dir .boardroom/documents
```

Each command prints a saved evidence ID and the extracted chunks. Add `--json` for structured output.

## Cite a page or block

Use the evidence ID to reopen an exact physical PDF page or a saved DOCX block:

```sh
node dist/cli.js evidence --id <pdf-evidence-id> --page 2 --data-dir .boardroom/documents
node dist/cli.js evidence --id <docx-evidence-id> --block 2 --data-dir .boardroom/documents
```

Exactly one of `--page` or `--block` is required with `--id`.
**What extraction does not do**
  Textual PDFs and DOCX text only: no OCR, and no preserved layout, images or formatting. DOCX blocks are saved extraction blocks, not Word page numbers. A failed extraction exits with code `2` and explains why; a partial one keeps its warnings.

On a packaged Windows candidate, replace `node dist/cli.js` with `.\boardroom.cmd`.