| name | find-wikipedia-pois |
| description | search Wikipedia for geotagged articles near a location and create POIs from the results. invoke when the user wants to find new points of interest for a city using Wikipedia |
| argument-hint | <content-path, e.g. europe/italy/lazio/rome> |
Search Wikipedia for geotagged articles near a World66 location and interactively review them as potential POIs.
Steps
1. Determine the location
If no argument is provided, ask the user which city they want to search. Resolve the content path and read the location's .md file to get its latitude and longitude.
2. Run the geosearch
Use tools/wiki_geosearch.py to find Wikipedia articles near the location. Search multiple points across the city to get good coverage:
python3 tools/wiki_geosearch.py LAT LNG --radius 10000 --json
For larger cities, search multiple centre points (downtown, major neighbourhoods) and deduplicate by pageid. The API returns max 50 results per query, so multiple queries from different points gives better coverage.
3. Filter results
Remove results that are not travel-relevant:
- Transit stations, bus stops, rail lines
- Residential buildings, condos, apartment complexes, office towers
- Schools, hospitals, fire stations, police departments
- Courts, government offices, detention centres
- Census-designated places, unincorporated areas
- Generic neighbourhood/district articles that duplicate what we already cover
Also remove any POIs the city already has (compare against existing .md files in the city's content directory).
4. Present candidates to the user
Show the filtered list grouped by likely category (museums, architecture, sights, restaurants, etc.). For each candidate show:
- Title
- Distance from city centre
- First sentence from Wikipedia
- Suggested category tag
Let the user review and select which ones to add. Go through them one by one or let the user pick from the list.
5. Create POI files
For each selected candidate:
- Fetch the Wikipedia article using WebFetch to get enough detail to write a good POI description
- Use the coordinates returned by the geosearch tool — do NOT make up or estimate coordinates. The geosearch results include
lat and lon fields; use those values rounded to 4 decimal places
- Write the POI file following LOCATIONS.md guidelines:
- Use proper tags: section tag (
things_to_do, eating_out, etc.), neighbourhood slug if applicable, category tag (museum, sight, architecture, etc.), and descriptive tags
- Set
neighbourhood: frontmatter for display if the POI is within a known neighbourhood
- Add a
story: field if you find a good anecdote in the Wikipedia article
- Write the description in World66 style (see STYLE.md) — not a Wikipedia summary, but a travel-guide perspective on why someone should visit
6. Commit
Commit the new POIs with a message like <city>: add N POIs from Wikipedia geosearch.
Important
- Always use Wikipedia's coordinates. The geosearch API returns precise lat/lon for each article. Never substitute your own guesses — they will often be hundreds of meters or even kilometres off.
- Read LOCATIONS.md before creating POIs, especially the tag system documentation.
- Neighbourhood POIs should only carry
things_to_do + neighbourhood as tags. Descriptive tags go on POIs within the neighbourhood, not on the neighbourhood itself.
- Quality over quantity. A city with 15 well-written POIs is better than 50 stubs. Only add places that a traveller would actually want to know about.
- Don't duplicate. Check what POIs the city already has before adding new ones.
