# DMS

DMS is a single-user document management system with:
- drag-and-drop upload anywhere in the UI
- file and folder upload
- document processing and indexing (PDF, text, OpenAI handwriting/image transcription, DOCX, XLSX, ZIP extraction)
- fallback handling for unsupported formats
- original file preservation and download
- metadata-based and full-text search
- learning-based routing suggestions

## Requirements

- Docker Engine with Docker Compose plugin
- Internet access for the first image build

## Install And Run With Docker Compose

1. Open a terminal in this repository root.
2. Start the full stack:

```bash
docker compose up --build -d
```

3. Open the applications:
- Frontend: `http://localhost:5173`
- Backend API docs: `http://localhost:8000/docs`
- Health check: `http://localhost:8000/api/v1/health`

## Setup

1. Open the frontend and upload files or folders.
2. Set default destination path and tags before upload if needed.
3. Configure handwriting transcription settings in the UI:
- OpenAI compatible base URL
- model (default: `gpt-4.1-mini`)
- API key and timeout
4. Open a document in the details panel, adjust destination and tags, then save.
5. Keep `Learn from this routing decision` enabled to train future routing suggestions.

## Data Persistence On Host

All runtime data is stored on the host using bind mounts.

Default host location:
- `./data/postgres`
- `./data/redis`
- `./data/storage`

To persist under another host directory, set `DCM_DATA_DIR`:

```bash
DCM_DATA_DIR=/data docker compose up --build -d
```

This will place runtime data under `/data` on the host.

## Common Commands

Start:

```bash
docker compose up --build -d
```

Stop:

```bash
docker compose down
```

View logs:

```bash
docker compose logs -f
```

Rebuild a clean stack while keeping persisted data:

```bash
docker compose down
docker compose up --build -d
```

Reset all persisted runtime data:

```bash
docker compose down
rm -rf ./data
```

## Handwriting Transcription Notes

- Handwriting and image transcription uses an OpenAI compatible vision endpoint.
- Before transcription, images are normalized:
  - EXIF rotation is corrected
  - long side is resized to a maximum of 2000px
  - image is sent as a base64 data URL payload
- Handwriting provider settings are persisted in host storage and survive container restarts.

## API Overview

GET endpoints:
- `GET /api/v1/health`
- `GET /api/v1/documents`
- `GET /api/v1/documents/{document_id}`
- `GET /api/v1/documents/{document_id}/preview`
- `GET /api/v1/documents/{document_id}/download`
- `GET /api/v1/documents/tags`
- `GET /api/v1/search?query=...`
- `GET /api/v1/settings`

Additional endpoints used by the UI:
- `POST /api/v1/documents/upload`
- `PATCH /api/v1/documents/{document_id}`
- `POST /api/v1/documents/{document_id}/reprocess`
- `PATCH /api/v1/settings/handwriting`