Wildside Recommendation Engine

Walking tours, powered by open data.

The Wildside engine generates personalized walking tours by combining OpenStreetMap geospatial data with Wikidata semantic information. Give it a starting point, a time budget, and your interests—it'll find the most rewarding route for you.

At its heart, the engine solves the Orienteering Problem: selecting which points of interest to visit (and in what order) to maximize your enjoyment within the time you have. Think of it as a very enthusiastic local guide who's read every Wikipedia article and knows exactly what you'd love to see.

Status

We're actively building this! The data foundation and scoring layers are in place; the route optimization solver is next on the list. See docs/roadmap.md for the full picture.

Crate overview

The engine is organized as a Cargo workspace with focused, single-purpose crates:

Crate	What it does
wildside-core	Domain model, traits, and abstractions—the shared vocabulary
wildside-data	ETL pipeline for ingesting OSM and Wikidata dumps
wildside-scorer	Popularity and user-relevance scoring
wildside-cli	Command-line tool for running the ingestion pipeline
wildside-fs	Filesystem abstraction layer

Quick start

You'll need the Rust toolchain (see rust-toolchain.toml for the exact version).

# Build the workspace
make build

# Run the test suite
make test

# Check formatting and lints
make check-fmt && make lint

To ingest data and build the POI database:

cargo run -p wildside-cli -- ingest \
  --osm-pbf path/to/region.osm.pbf \
  --wikidata-dump path/to/wikidata-dump.json.bz2 \
  --output-dir ./data

This produces pois.db (SQLite database), pois.rstar (spatial index), and popularity.bin (precomputed scores)—the artefacts consumed at runtime.

Documentation

For API details, usage patterns, and integration guidance, see the User Guide. The Design Document covers architectural decisions and implementation rationale.

Licence

ISC