Workshop Walker is an alternate interface to the Steam Workshop, making it easier to find the items you care about.
It does this by taking the data freely provided by Valve, analysing it and serving it up in a tidy package.
Specific improvements:
- Dependant discovery, find everything that depends on this item.
- Language detection, filter for specific languages. No more poorly marked translations.
- Properties, extended classification, submit new ones or vote on existing entries for relevance.
Interested? You can check it out live at https://workshop-walker.disconsented.com/
Workshop Walker pulls every item from the Steam Workshop periodically via their API, processing each item individually for the following:
- Language detection on the item's body.
- Converting BB Code to HTML for rendering.
- ML assessments for properties if the description has changed.
- Inserting relationships between items. This is something that SurrealDB lets us model cleanly.
This project was created for RimWorld originally, however it is designed to be agnostic and work with any Steam Workshop game.
If you think that it'd be useful to include, please open a new discussion.
Please do take care to explain how Workshop Walker will be useful, I'd rather not add games that won't benefit or won't get used.
Want to contribute a feature? Great!
Please suggest it as an issue to get a feeling on whether I'll accept it. Remember, that every piece of code contributed adds to the maintenance burden. This is also an effort to manage burnout.
For anything touching the UI, ideally include some kind of visual reference.
If it's accepted:
- Commit messages follow Conventional Commits. release-plz builds the changelog and version from them.
- Rust:
cargo +nightly fmt --allandcargo clippy --allmust pass. Clippy warnings fail the build. - UI:
npm run format,npm run lintandnpm run checkfromui/must pass.
- Rust, latest stable
- Clang, to build RocksDB (
libclang-devon Debian and Ubuntu) - Node.js 26
- biscuit-cli, to make a keypair
- A Steam Web API key
- a llama.cpp server for property classification
The database is embedded. It writes to ./workshopdb and needs no server.
git clone https://github.com/disconsented/workshop-walker
cd workshop-walker
cp config/config.example.toml config/config.toml
biscuit keypair Put the keypair, your Steam API key and your SteamID64 in config/config.toml.
To run without classification, set ml_extraction.enabled = false.
docker run -d -p 8081:8080 \
--volume models:/models --env LLAMA_CACHE=/models/ \
--restart unless-stopped \
ghcr.io/ggml-org/llama.cpp:server \
-hf google/gemma-4-E4B-it-qat-q4_0-gguf -hff gemma-4-E4B_q4_0-it.gguf \
-c 4096 -np 1 -t 20 -tb 20 --jinja --reasoning off \
--sleep-idle-seconds 120 --no-warmup Backend (listens on port 5800):
RUST_LOG="workshop_walker=debug,salvo=debug" cargo run --bin workshop-walkerFrontend, in a second terminal:
cd ui
npm install
npm run devOpen http://localhost:5173. The dev server sends /api to the backend.
This project is not associated with Valve, Steam, etc. I just got annoyed one day I couldn't see dependants and wanted an excuse to use SurrealDB relations.