pi-estate-agent

A Pi package that turns Pi into a Danish real-estate agent for boligsiden.dk — it finds houses matching your criteria, compares them, and keeps stored listings up to date.

The agent works through three skills backed by native Pi tools that talk to Boligsiden and other Danish public-data services. Promising listings (and their photos) are saved to a local database so you can compare them and track price changes over time.

Skills

Skill	What it does
/skill:search	Search Boligsiden by location, price, size and type; judge suitability; store good listings and their photos.
/skill:compare	Compare stored listings on economy, living space, and commute/travel.
/skill:update	Re-check stored listings for changes (price, days on market, availability).

Integrations

• Boligsiden — houses for sale (search + case details)
• Skraafoto — oblique aerial 360 photos (Dataforsyningen STAC API)
• Rejseplanen — public-transport journey planning
• DAWA — Danish address lookup (text → id + coordinates)

Storage uses libsql (@libsql/client, with vector-search support for future use). Listing photos are downloaded to ./data/images/.

Requirements

• Pi (npm install -g @earendil-works/pi-coding-agent)
• Node.js 18+ (bundled with Pi's installer)

Install

As a Pi package (recommended)

Install it into Pi from git; Pi runs npm install for the extension's dependencies automatically:

pi install git:github.com/mrn-dk/pi-estate-agent

Use -l to install it only for the current project instead of globally. Once installed, start pi and the estate_* tools and /skill:* skills are available in any session.

From a local clone (development)

git clone https://github.com/mrn-dk/pi-estate-agent
cd pi-estate-agent
npm install            # installs @libsql/client used by the extension
pi install -l .        # register this repo as a project-local Pi package

Then run pi in the project directory (trust it when prompted). The database is created automatically under ./data/ on first use.

Usage

Invoke a skill with /skill:<name>, optionally followed by instructions. You can also just describe what you want in plain language — Pi loads the right skill and tools on its own.

/skill:search find a villa near Aarhus under 4.5M kr with at least 4 rooms and a garden

A typical session:

1. Search & save — /skill:search looks up matching homes, judges them against your criteria, and stores the good ones with their photos:

   /skill:search condos in Copenhagen (zip 2100) between 3 and 5 million, min 3 rooms
   

2. Compare — /skill:compare weighs the saved listings against each other on economy, living space, and commute:

   /skill:compare the two cheapest saved listings, commute target Rådhuspladsen 1 København
   

3. Update — /skill:update re-checks saved listings for price drops or changes:

   /skill:update all saved listings
   

Data is stored under ./data/ (git-ignored): the libsql database in data/estate.db and downloaded photos in data/images/<caseID>/. Override the location with ESTATE_DATA_DIR or ESTATE_DB.

Tools

All integrations are native Pi tools (no CLI to run):

Tool	Purpose
estate_search	Search Boligsiden listings by location/price/size/type
estate_case	Fetch one listing by caseID / URL / slug
estate_address_search	Resolve an address to id + coordinates (DAWA)
estate_skraafoto	Oblique aerial 360 photos for an address / lat,lon
estate_transit_locations / estate_transit_trip	Rejseplanen journeys
estate_db_add / estate_db_list / estate_db_get / estate_db_update / estate_db_remove	Stored-listing database

See skills/TOOLS.md for parameters and details.

Notes

• DinGeo is not integrated: it sits behind a Cloudflare challenge that blocks programmatic access.
• Listing photos are stored but not analysed by a vision model in this version.