Updated docs

This commit is contained in:
2026-08-11 23:55:20 +00:00
parent 550cb841ec
commit 023d224179
2 changed files with 61 additions and 46 deletions
+60 -45
View File
@@ -1,77 +1,92 @@
# mediapreview
Low-level media preview converters plus an optional persistent worker pool
framework. All converters produce AVIF output.
Generate compact AVIF preview images from images, videos, PDFs and office documents.
## Layout
`mediapreview` is a small library of low-level converters. Give it a file path and
it returns AVIF bytes plus a response object telling you which backend handled it
and whether it succeeded.
| Module | Purpose |
|--------|---------|
| `mediapreview.backends.image` | Image → AVIF via pyvips (ffmpeg for HEIC/HEIF/AVIF) |
| `mediapreview.backends.video` | Video frame → AVIF via PyAV (HDR preserved) |
| `mediapreview.backends.pdf` | PDF/XPS/EPUB page → AVIF via PyMuPDF + pyvips |
| `mediapreview.backends` | `dispatch()` — pick a backend by path/mimetype |
| `mediapreview.formats` | Suffix sets, previewability and priority classification |
| `mediapreview.office` | OnlyOffice Document Server client + Docker bootstrap |
| `mediapreview.docker/` | Patched OnlyOffice image build context (ships in the wheel) |
| `mediapreview.pool` | Async persistent subprocess worker pool (optional) |
| `mediapreview.worker` | Worker subprocess entry point (framed stdin/stdout protocol) |
| `mediapreview.protocol` | msgspec wire structs (`PreviewRequest` / `PreviewResponse`) |
| `mediapreview.cache` | Thread-safe LRU cache for preview responses |
## Install
## Extras
```bash
pip install mediapreview # image converter only (pyvips)
pip install mediapreview[pdf] # + PDF/XPS/EPUB (pymupdf)
pip install mediapreview[video] # + video (av, numpy)
pip install mediapreview[office] # + OnlyOffice client (httpx, pyjwt)
pip install mediapreview[worker] # + worker pool (blake3, tracerite)
pip install mediapreview[standard] # everything
```sh
uv add mediapreview[standard]
```
## Usage
`[standard]` pulls in every backend and the optional worker pool. Use the feature-specific extras to avoid dependencies not needed for your application.
Low-level, in-process:
## Quick start
```python
from pathlib import Path
from mediapreview import dispatch
avif_bytes, resp = dispatch(path, quality=60, maxsize=512, maxzoom=2.0)
avif_bytes, resp = dispatch(
Path("photo.jpg"),
quality=60,
maxsize=512,
maxzoom=2.0,
)
if resp.ok:
Path("preview.avif").write_bytes(avif_bytes)
else:
print(resp.error)
```
Worker pool (isolates heavy imports and native crashes from the async loop):
`dispatch` picks the backend from the file extension or mimetype. You can also
call the backend functions directly:
```python
from mediapreview.pool import start_preview_workers, shutdown_preview_workers
from mediapreview import process_image, process_pdf, process_video
avif, resp = process_image(Path("photo.jpg"), maxsize=512, quality=60)
avif, resp = process_pdf(Path("doc.pdf"), maxsize=512, quality=60, page_number=0)
avif, resp = process_video(Path("clip.mp4"), maxsize=512, quality=60)
```
One-shot CLI (works with the base install, no worker extra needed):
## Worker pool (optional)
Heavy native dependencies and crashes stay out of your async loop by running
previews in a pool of persistent subprocess workers.
```python
from mediapreview.pool import (
start_preview_workers,
shutdown_preview_workers,
run_preview,
)
await start_preview_workers()
try:
avif, resp = await run_preview(Path("doc.pdf"))
finally:
await shutdown_preview_workers()
```
Add the `worker` extra to use the pool.
## CLI
```bash
mediapreview photo.jpg -o preview.avif # or: python -m mediapreview ...
mediapreview doc.pdf -q 70 --maxsize 1024 # needs the matching backend extra
mediapreview photo.jpg -o preview.avif
mediapreview doc.pdf -q 70 --maxsize 1024
mediapreview oosetup # build + run the bundled OnlyOffice container
```
## OnlyOffice Docker bootstrap
## OnlyOffice setup
A patched OnlyOffice image (configurable converter worker count) ships as
package data and can be built/started with:
Office previews need an OnlyOffice Document Server. A patched Docker image ships
inside the package and can be started with:
```bash
mediapreview oosetup # builds + runs "onlyoffice-mediapreview" on port 8988
mediapreview oosetup [name] [port]
```
Container name and port are optional positional args:
`mediapreview oosetup [name] [port]`.
`oosetup` logs progress to stderr and prints exactly one line on stdout:
`oosetup` logs to stderr and prints one line on stdout:
```
ONLYOFFICE_JWT_SECRET=<token>
```
If `ONLYOFFICE_JWT_SECRET` is already set in the environment it is used as-is
(and echoed back); otherwise a random secret is generated. Persist the token
wherever your deployment keeps its configuration and export it for later runs
— the caller owns the secret, mediapreview does not store it.
Set `ONLYOFFICE_JWT_SECRET` yourself to reuse an existing secret; otherwise a
random one is generated.
+1 -1
View File
@@ -5,7 +5,7 @@ build-backend = "hatchling.build"
[project]
name = "mediapreview"
dynamic = ["version"]
description = "Low-level media preview converters and a worker pool framework"
description = "Fast media preview images of photos, videos, PDFs and office documents."
readme = "README.md"
requires-python = ">=3.11"
dependencies = [