# Reference Documents

Material that is **about you but is not a project**: your CV, an about page, a
FAQ, a talk abstract. Everything here is indexed and retrievable, and never
announced in the project catalogue.

Accepts `.md` and `.mdx`.

## Why this directory exists

`data/projects/` is advertised. The bot injects the full list of what lives
there into every prompt, so visitors get told those are your projects.

Your CV is the document that answers what someone considering hiring you
actually asks — "does he know Kubernetes?", "where has he worked?", "how long
was he at that job?" — and none of it is retrievable while it lives only in
your site. But it is not a project, and putting it in `data/projects/` makes
the bot cheerfully list "cv" as one of your works. Hence the split.

## The contents are gitignored

Only this README is tracked. What goes here is personal (a CV) or a symlink to
a path that only exists on your machine, and neither travels well in a repo.
Set it up on each install:

```bash
ln -s ../../../portfolio/src/content/cv/cv.mdx data/docs/cv.mdx
./bin/chat-bot reindex
```

A symlink rather than a copy so there is one file to keep current — edit the
CV in your site, re-run `reindex`, and the bot is up to date.

## Configuration

```yaml
rag:
  data_path: ./data/projects   # projects → listed in the catalogue
  docs_path: ./data/docs       # this directory → retrievable, never listed
```

Leave `docs_path` empty to turn the whole thing off.

## Notes

- Frontmatter is excluded from retrieval. It is dense metadata in a very short
  chunk, which makes it a magnet for short queries — a CV's `location:` field
  was answering *"where has Victor worked?"* with a city instead of his work
  history.
- Long sections split at `###` headings, so each job in a CV's Experience
  section stays one chunk instead of being cut mid-sentence.
- This README is skipped by the indexer, as is the one in `data/projects/`.