Hsl

HSL / Digitransit Transit Skill

Query HSL public transport data using Digitransit APIs.

This skill is designed for:

• stop and station lookup
• next departures from a stop
• route planning between two places in the HSL area

API Notes

The official HSL routing backend is Digitransit GraphQL:

• Endpoint: https://api.digitransit.fi/routing/v2/hsl/gtfs/v1
• Method: POST
• Content type: application/json or application/graphql
• Registration and API keys are required by Digitransit docs

At implementation time in this environment, both direct unauthenticated Digitransit GraphQL routing requests and geocoding requests returned HTTP 401. The official auth model appears to be API-key based via the Digitransit portal, not username/password Basic auth. This skill now tries subscription-key style headers and becomes live once a valid API key is configured.

Tools

scripts/lookup_places.py

Resolve place names, addresses, and stop names in the HSL area.

./scripts/lookup_places.py Kamppi
./scripts/lookup_places.py "Pasila asema" --limit 8
./scripts/lookup_places.py Tapiola --json

Use this when:

• the user mentions a stop, station, address, district, or landmark
• you need coordinates before route planning
• you need candidate HSL stops to disambiguate a place name

scripts/query_routes.py

Live HSL route planning between two places.

./scripts/query_routes.py --from "Kamppi" --to "Tapiola"
./scripts/query_routes.py --from "Hakaniemi" --to "Lentoasema" --arrive-by
./scripts/query_routes.py --from "Pasila" --to "Otaniemi" --json

This script:

• resolves route endpoints via geocoding or built-in HSL place fallbacks
• calls Digitransit GraphQL live
• prints human-friendly itineraries by default
• emits full payload + response with --json

Use this when:

• the user asks for an HSL route from A to B
• you want live itinerary options with real-time delay info
• you need raw Digitransit output for debugging with --json

scripts/next_departures.py

Live next departures from an HSL stop or station.

./scripts/next_departures.py Kamppi
./scripts/next_departures.py "Tapiola" --departures 8 --mode metro
./scripts/next_departures.py "Pasila" --stop-id HSL:1040417 --json

This script:

• queries HSL stop departures live from Digitransit GraphQL
• picks the best-matching stop for a free-text stop name
• supports --mode to prefer bus, tram, rail, metro, or ferry stops
• supports --stop-id for an exact GTFS stop match
• prints human-friendly departures by default
• emits raw stop data with --json

Use this when:

• the user asks for next departures from a stop or station
• the user wants tram, bus, metro, ferry, or local train departures at a named HSL stop
• you need the stop code / GTFS ID alongside departures

Use lookup_places.py first if the stop name is ambiguous.

Use this when:

• the user asks for an HSL route from A to B
• you want normalized coordinates and a ready-to-send routing query

Agent Instructions

When the user asks about HSL transit:

1. Resolve ambiguous place names with lookup_places.py
2. If the user wants a route, run query_routes.py
3. If the user wants departures from a stop or station, run next_departures.py
4. Be explicit when a result is based on fallback place resolution versus live Digitransit routing or departure data

Known limitations

• Free-text stop matching for departures is best-effort and may still need lookup_places.py first for ambiguous names.
• Built-in fallback coordinates exist only for a small set of common places.

Confirmed implementation notes

From the official Digitransit docs and live probing:

• HSL routing GraphQL endpoint: https://api.digitransit.fi/routing/v2/hsl/gtfs/v1
• Request method: POST
• Direct unauthenticated routing requests returned 401 Access Denied
• Digitransit docs state registration is required
• Geocoding is the right companion API for resolving free-text places before route planning

Suggested GraphQL operations

The skill scaffold is built around these likely Digitransit GraphQL operations:

• plan(from: ..., to: ..., numItineraries: ...)
• stops(name: ...)
• stop(id: ...)
• routes(name: ...)
• places(name: ...)

These are included as implementation targets in references/api.md and in the route query scaffold.