RSSBox: per-source translation engines over a Django feed pipeline
Smarter RSS management, more efficient information access
At a glance
- What is it?
- RSSBox is a Django application that translates RSS titles, bodies and summaries with DeepL or any OpenAI-compatible model, serves the result back as RSS or JSON, and merges tagged sources into one filtered feed. The dependency list also puts the debug toolbar in production.
- Who is it for?
- RSSBox fits a reader who follows a lot of non-English feeds and wants one subscription list instead of one translation toggle per site, because the per-source engine choice and the tag merge are the two features other setups tend to leave you. It does not fit a minimal deployment, since the runtime dependency list carries development tooling and the container expects cron and a writable data directory.
- Can I use it commercially?
- Yes, with conditions. GPL-3.0 is a copyleft licence: if you distribute software that includes it, you must release that software's source code under the same licence. Running it internally without distributing it does not trigger that obligation.
- Is it still maintained?
- Yes. The repository received new commits within the last day.
- What is it written in?
- Mainly Python, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 22, 2026, and from our analysis. They are not legal advice.
Star history
- versun/RSSBox683 ★ · +6
| Repository | 2026-09-11 | 2026-11-22 |
|---|---|---|
| versun/RSSBox | 677 | 683 |
6 Stars in the last 72 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
The feature list jumps from 5 to 7 and from 7 to 9
The feature list in the README is numbered 1, 2, 3, 4, 5, 7, 9, 10 and 11. Two numbers are missing, which reads like two features that were pulled without renumbering the rest. The contribution guide has the same shape, numbered 1 through 4 and then 6.
What survives the numbering still describes a specific product rather than a translator. It translates the title, the content or the summary, shows two languages side by side, and lets you subscribe to the translated feed as RSS or JSON. Beyond that it controls the update frequency of each translation source and shows the translation status, reports the token and character spend per source, fetches full text, filters by keyword or by AI, and merges several sources into one unified source through tags with filters applied.
So the unit of work is the source, not the feed. Each source carries its own engine, its own schedule, its own status and its own cost line, and tags are what let a group of sources behave like one.
The Docker commands still build the old image name
The project was renamed. The README subtitle still says it was originally RSS Translator, and the home page is rssbox.app, but the comment block at the top of the Dockerfile builds both identities:
# Build in local: docker build . --platform linux/arm64 -t versun/rsstranslator:dev versun/rssbox:devThe multi-arch line in the same block pushes four tags, `versun/rsstranslator:latest`, `versun/rssbox:latest` and the two version tags, and it creates a buildx builder first. The file also refers to `docker-compose.test.yml` and `docker-compose.dev.yml`, neither of which sits in the repository root listing.
That is the practical consequence of the rename: a self-hoster who copies a build line from the Dockerfile can end up pushing an image called rsstranslator while the application calls itself rssbox, and anyone following an older deployment guide will look for a compose file that is no longer at the root.
The debug toolbar and coverage sit in the runtime dependencies
One flat dependency list carries twenty-six entries, and every constraint is a floor with no ceiling. The runtime side is recognisable: Django, redis, the encrypted model fields package, django-tagulous for the tag merge, bs4, lxml, mistune, html2text, gunicorn and whitenoise. Two entries explain two features directly. `newspaper4k` is a full text extractor, which is the get-full-text item, and `tiktoken` is a token counter, which is the per-source token and character cost line.
The two entries that do not belong are `django-debug-toolbar` and `coverage`. Both are development tools, and both sit in the same list as gunicorn, so a production install resolves them whether or not anyone opens a debug page. The `[tool.coverage.report]` block also carries exclusions for `if settings.DEBUG` and `if settings.DEMO`, which tells you the demo mode is a settings switch that the tests deliberately skip.
Three console entry points decide how the app starts
The manifest declares three console scripts rather than one. `server` maps to `scripts.prod_server:main`, `dev` maps to `scripts.dev_server:main`, and `init` maps to `scripts.init:init_server`, which is the initialisation step the contribution guide mentions in passing.
Development is a clone, a branch name and one command:
git clone https://github.com/versun/rssbox.git
cd rssboxuv run dev # 将会自动安装依赖包并完成初始化Because the uv section sets `package = true`, that single invocation installs the dependencies and initialises the project. The same file also keeps a setuptools packages list beside the uv configuration, naming config, core, locale, scripts, templates and utils, so the templates directory is treated as a distributable package while deploy, docs and the development diaries are left out. The coverage omissions agree, dropping scripts, manage.py, migrations and the three config entry points.
The image ships cron, gosu and a cleaned virtualenv
The Dockerfile is a two stage build on `python:3.13-slim-bookworm`. The builder installs build-essential and git, creates a virtualenv at `/opt/venv`, installs uv and runs `uv pip install --no-cache-dir -e .`, so the dependency resolution happens once at image build time rather than at first request.
The final stage is where the operational detail sits. It copies the virtualenv and the application out of the builder, installs gosu, cron and gettext, sets `PORT=8000` and `PYTHONPATH=/app`, creates `/var/run/cron` with a `crond.pid`, and copies a cron configuration in from `config/rt_cron`. Cron inside the image is what makes per-source update schedules work, so a schedulerless host will not reproduce the documented behaviour. The stage also walks the copied virtualenv and deletes every `__pycache__` directory and `.pyc` file, a size pass that runs on the installed tree rather than on the source.
Feeds are versioned by date, and three commits sit past the tag
Releases are named for the day they shipped: 2026.9.16, 2026.9.6 and 2026.9.5. The project version in the manifest reads 2026.9.16, so the manifest and the newest tag agree. The last push is 2026-09-19, three days after that tag, which means the branch carries work that no release has covered.
The runtime floor is Python 3.12, while the container is built on 3.13 and the repository keeps its own `.python-version` file, so three places express the version question and only one of them is a hard error if you get it wrong.
The Python requirement is also why the library set looks the way it does. gevent and gunicorn sit side by side for the worker story, `feedparser`, `feed2json` and `feedgen` cover the read and write ends of the feed format, and `python-dateutil`, `httpx` and `fake-useragent` handle the messy edges of third party sources. That set is what turns a reader into a publisher, which is the whole point of the subscribe-to-translated-output feature.
A development diary and a public ledger ship with the code
The root listing includes a `dev-diaries/` directory, so the working notes are part of the repository rather than a separate blog. It also includes `SECURITY.md`, `deploy/` and `locale/`, and the README links out to a Telegram group, a GitHub projects board for development progress and a Notion page that is an income and expense sheet for the project.
The sponsorship block names Railway as an open source supporter and then points at two personal sponsorship routes, with a note that sponsor avatars refresh at 05:30 UTC twice a day. The README also carries a star history chart link and a DeepWiki link.
None of that is unusual for a small self-hosted project, but it does tell you something about the shape of the repository: the changelog is a diary, the roadmap is a project board, the money is public, and the only versioned history is a date in the tag name.
Editorial conclusion
RSSBox fits a reader who follows a lot of non-English feeds and wants one subscription list instead of one translation toggle per site, because the per-source engine choice and the tag merge are the two features other setups tend to leave you. It does not fit a minimal deployment, since the runtime dependency list carries development tooling and the container expects cron and a writable data directory. Before you commit, read the floor-only version constraints, decide whether DeepL or a local OpenAI-compatible endpoint handles your languages, and remember that the project renamed itself from RSS Translator while its Docker commands still build the old image name.
Frequently asked questions
What does RSSBox do to an RSS feed?
It translates the title, the content or the summary, can display both languages, and lets you subscribe to the translated output as RSS or JSON. Full text fetching, keyword filtering and AI filtering are also listed as features.
Which translation engines does RSSBox support?
DeepL and any model behind an OpenAI-compatible interface. Each source can be assigned its own engine, its update frequency can be controlled, and the token and character cost per source is visible.
How do I run RSSBox for development?
Clone the repository and run `uv run dev`, which installs the dependencies and completes initialisation. Before opening a pull request the project asks that the ruff format check and the basic tests pass.
What Python version does RSSBox need?
3.12 or newer, declared as `requires-python = ">=3.12"`. The container image is built on python:3.13-slim-bookworm and the repository also carries a .python-version file.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/versun-rssbox)Star history chart for your README
[](https://hysenlabs.com/projects/versun-rssbox)Was this useful?