# Release notes

What changed in each version of Curator, newest first. Curator updates itself, so you do not need to download anything. See [Updates and versions](/reference/updates/) for how updates arrive.

## 0.0.2

- Curator now runs its AI features through Ollama, LM Studio, or any OpenAI-compatible endpoint. Onboarding detects installed engines, can install one for you, and recommends models by tier.
- Processing is now cost aware. OCR and image description wait behind an explicit heavy run, so you decide when to spend the time and compute.
- CSV and TSV files are processed with real structure extraction plus a short description.
- Smarter PDF handling. Scanned PDFs get a quick vision check before OCR, garbled text is no longer stored as content, and you can request an AI description for any PDF.
- Files that fail to process now show why, offer a retry, and link to the setting that fixes it.
- Refreshed interface with reworked buttons, switches, checkboxes, dialogs, progress bars, icons, and clearer search result labels.
- Organize runs now survive navigating away and no longer start twice, and chat no longer hijacks scrolling.

## 0.0.1

The first public release.

- Browse files from a local folder or an S3 bucket in one native macOS window.
- On-device processing that reads your files and builds a searchable index, with your own local models through Ollama.
- Semantic search that finds files by meaning, not just by filename.
- Chat with an assistant that can look across your files and answer questions about them.
- Auto-organization that proposes a folder structure and applies it for you.