OpenStreetMap • Open WebUI Community

""" title: OpenStreetMap Tool author: projectmoon author_url: https://git.agnos.is/projectmoon/open-webui-filters version: 4.1.2 license: AGPL-3.0+ required_open_webui_version: 0.6.31 requirements: openrouteservice, pygments """ import itertools import json import math import requests from fastapi.responses import HTMLResponse from pygments import highlight from pygments.lexers import JsonLexer from pygments.formatters import HtmlFormatter import openrouteservice from openrouteservice.directions import directions as ors_directions from urllib.parse import urljoin from operator import itemgetter from typing import List, Optional, Tuple, Literal, TypeAlias from pydantic import BaseModel, Field ##################################################### # Map UI Integration ##################################################### NWR: TypeAlias = Literal['node', 'way', 'relation'] POICategory: TypeAlias = Literal[ 'Groceries', 'Recreation', 'Restaurants & Bars', 'Travel & Tourism', 'Healthcare', 'Education & Schools', 'Fuel & EV Charging', 'Religious Locations' ] class OsmPOI(BaseModel): """Represents a Point of Interest found during map search.""" name: str = Field(description="Name of the Point of Interest.") address: Optional[str] = Field(description="Address of the Point of Interest.") description: str = Field(description=("Useful one sentence description of the Point of Interest, " "containing important or unique information. " "Do not put distance here.")) distance: float = Field(description="Distance in km or miles from the search location.") distance_unit: Literal["km", "mi", "kilometers", "miles"] = Field(description="Unit of distance.") lat: float = Field(description="Latitude of the Point of Interest.") lon: float = Field(description="Longitude of the Point of Interest.") osm_id: str = Field(description="OpenStreetMap ID of the Point of Interest.") osm_type: NWR = Field(description="OpenStreetMap entity type of the Point of Interest.") pass MAP_UI = """ <!DOCTYPE html> <html lang="en"> <head> <meta charset="UTF-8"> <meta name="viewport" content="width=device-width, initial-scale=1.0"> <meta name="color-scheme" content="light dark"> <title>Search Results</title> <link rel="stylesheet" href="https://unpkg.com/leaflet@1.9.4/dist/leaflet.css" /> <style> * { margin: 0; padding: 0; box-sizing: border-box; } body { font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif; color: #333; line-height: 1.6; padding: 12px; } .container { max-width: 1200px; margin: 0 auto; } .results { display: flex; flex-wrap: wrap; align-items: start; justify-content: space-between; } @media (width <= 600px) { .results { flex-direction: column; gap: 50px; } .map-container { flex: auto; min-width: 100%; } } @media (width > 600px) { .results { flex-direction: row; gap: 50px; } .map-container { flex: auto; } } h1 { font-size: 2rem; font-weight: 700; color: #2c3e50; text-align: center; margin-bottom: 5px; } .subtitle { font-size: 0.9rem; color: #6c757d; text-align: center; margin-bottom: 15px; } h2 { font-size: 1.3rem; color: #34495e; margin: 20px 0 10px; font-weight: 600; } #map { height: 400px; margin-bottom: 15px; border-radius: 8px; box-shadow: 0 4px 12px rgba(0, 0, 0, 0.1); border: 1px solid #ddd; } .poi-list-container { flex: initial; } #poi-list { list-style-type: none; max-height: 400px; overflow-y: auto; } #poi-list li { padding: 10px 12px; margin: 0; background-color: white; border-radius: 6px; cursor: pointer; transition: all 0.2s ease; box-shadow: 0 2px 4px rgba(0, 0, 0, 0.05); border: 1px solid #e9ecef; font-weight: 600; color: #495057; display: flex; flex-direction: column; } #poi-list li:hover { transform: translateY(-2px); box-shadow: 0 3px 8px rgba(0, 0, 0, 0.1); background-color: #f8f9fa; color: #2c3e50; } #poi-list li.active { background-color: #3498db; color: white; box-shadow: 0 3px 8px rgba(52, 152, 219, 0.2); } .poi-description { font-size: 0.8rem; color: #555; /* Slightly lighter than default */ margin-top: 4px; font-weight: 400; } .poi-distance { font-size: 0.8rem; color: #444; /* Slightly lighter than default */ margin-top: 4px; font-weight: 400; } #poi-list li.active .poi-description { color: rgba(255, 255, 255, 0.9); /* Lighter text for active state */ } </style> </head> <body> <div class="container"> <h1>Map Results for {REPLACE_WITH_CATEGORY}</h1> <p class="subtitle">Data from OpenStreetMap</p> <div class="results"> <div class="map-container"> <h2>Map</h2> <div id="map"></div> </div> <div class="poi-list-container"> <h2>POI List</h2> <ul id="poi-list"></ul> </div> </div> </div> <script src="https://unpkg.com/leaflet@1.9.4/dist/leaflet.js"></script> <script> // POI data with coordinates and names // [ { name, coords, lat, lon, description } ] const poiData = {REPLACE_WITH_POI_RESULTS}; // Function to create the map function createMap(poiData) { // Create map centered on first POI const map = L.map('map', { minZoom: 3, maxZoom: 18 }); if (poiData.length > 0) map.setView([poiData[0].lat, poiData[0].lon], 12); // Add OpenStreetMap tiles L.tileLayer('{REPLACE_WITH_TILE_SERVER}', { attribution: '© <a href="https://www.openstreetmap.org/copyright">OpenStreetMap</a> contributors' }).addTo(map); // Store markers and popup references const markers = []; // Add markers to map and create list items dynamically poiData.forEach((poi, index) => { const marker = L.marker([poi.lat, poi.lon]).addTo(map); const osmLink = `https://www.openstreetmap.org/${poi.osm_type}/${poi.osm_id}` const header = `<h3><a target="_blank" href="${osmLink}">${poi.name}</a></h3>` const address = `<p>${poi.address}</p>` const description = `<p>${poi.description}</p>` const coords = `<p>Coordinates: ${poi.lat.toFixed(4)}, ${poi.lon.toFixed(4)}</p>` const distance = `${poi.distance.toFixed(2)} ${poi.distance_unit}` marker.bindPopup(`${header}${address}${description}<p>${distance}</p>${coords}`); markers.push(marker); // Create list item dynamically const listItem = document.createElement('li'); listItem.innerHTML = ` <span>${poi.name}</span> <span class="poi-description">${poi.description}</span> <span class="poi-distance">${distance}</span>`; listItem.onclick = () => focusOnMarker(index, map, markers); listItem.classList.add('poi-item'); // Add to list document.getElementById('poi-list').appendChild(listItem); }); return { map, markers }; } // Function to focus on a specific marker function focusOnMarker(index, map, markers) { const marker = markers[index]; // Fit map to show all markers with reduced padding map.flyTo(marker.getLatLng(), 15); // Open popup for selected marker marker.openPopup(); // Update active state in list updateActiveListItem(index); } // Function to update active list item function updateActiveListItem(index) { const listItems = document.querySelectorAll('#poi-list li'); listItems.forEach((item, i) => { if (i === index) { item.classList.add('active'); } else { item.classList.remove('active'); } }); } // Initialize map and markers const { map, markers } = createMap(poiData); // Initialize with first POI selected updateActiveListItem(0); </script> </body> </html> """ ##################################################### # Category Type Aliases ##################################################### UrbanSetting: TypeAlias = Literal['urban', 'suburban', 'rural'] StoreCategory: TypeAlias = Literal[ 'groceries', 'convenience', 'alcohol', 'drugs', 'cannabis', 'electronics', 'electrical', 'hardware', 'diy' ] RecreationCategory: TypeAlias = Literal[ 'swimming', 'playgrounds', 'amusement', 'sports', 'gyms' ] EateryCategory: TypeAlias = Literal[ 'sit_down_restaurants', 'fast_food', 'cafe_or_bakery', 'bars_and_pubs' ] TravelCategory: TypeAlias = Literal[ 'tourist_attractions', 'accommodation', 'bike_rentals', 'car_rentals', 'public_transport' ] HealthcareCategory: TypeAlias = Literal[ 'doctor', 'dentist', 'hospital', 'pharmacy' ] EducationCategory: TypeAlias = Literal[ 'schools', 'universities_and_colleges', 'libraries' ] FuelCategory: TypeAlias = Literal['fossil_fuels', 'ev_fast_charging'] FuelTypeCategory: TypeAlias = Literal['petrol', 'diesel', 'all'] EVChargerCategory: TypeAlias = Literal[ 'chademo', 'chademo3', 'chaoji', 'ccs2', 'ccs1', 'gb/t', 'nacs', 'all' ] # Religions from the OSM wiki. Some make use of the denomination tag. # Others do not. any_religion is a special "category" that forces the # tool to search for any place of worship amenity. ReligionCategory: TypeAlias = Literal[ 'any_religion', 'buddhist', 'christian', 'hindu', 'jain', 'jewish', 'muslim', 'pagan', 'shinto', 'sikh', 'taoist', 'zoroastrian', 'ancestor', 'animist', 'antoinist', 'bahai', 'benzhu', 'caodaism', 'chinese_folk', 'confucian', 'shamanic', 'scientologist', 'self-realization_fellowship', 'spiritualist', 'tenrikyo', 'unitarian_universalist','vietnamese_folk', 'voodoo', 'yazidi', 'multifaith', 'humanist', 'atheist', 'null', 'laica' ] DenominationCategory: TypeAlias = Literal[ # Forcibly search for any denomination within the specified # religion. 'any_denomination', # Buddhist "buddhist|mahayana", "buddhist|gelug", "buddhist|jishu", "buddhist|jodo_shinshu", "buddhist|jodo_shu", "buddhist|nichiren", "buddhist|nyingma", "buddhist|obaku", "buddhist|pure_land", "buddhist|rinzai", "buddhist|risshu", "buddhist|shingon_shu", "buddhist|soto", "buddhist|tiantai", "buddhist|tibetan","buddhist|vajrayana", "buddhist|won", "buddhist|yogacara", "buddhist|yuzu_nembutsu", "buddhist|zen", "buddhist|theravada", "buddhist|thai_mahanikaya", "buddhist|thai_thammayut", # Christian "christian|catholic", "christian|armenian_catholic", "christian|chaldean_catholic", "christian|coptic_catholic", "christian|eritrean_catholic", "christian|ethiopian_catholic", "christian|greek_catholic", "christian|hungarian_greek_catholic", "christian|maronite", "christian|polish_catholic", "christian|roman_catholic", "christian|romanian_catholic", "christian|syriac_catholic", "christian|syro-malabar_catholic", "christian|ukrainian_greek_catholic", "christian|orthodox", "christian|antiochian_orthodox", "christian|armenian_apostolic", "christian|bulgarian_orthodox", "christian|eritrean_orthodox", "christian|ethiopian_orthodox", "christian|coptic_orthodox", "christian|georgian_orthodox", "christian|greek_orthodox", "christian|macedonian_orthodox", "christian|old_believers", "christian|polish_orthodox", "christian|romanian_orthodox", "christian|russian_orthodox", "christian|serbian_orthodox", "christian|syriac_orthodox", "christian|ukrainian_orthodox", "christian|protestant", "christian|adventist", "christian|anabaptist", "christian|anglican", "christian|baptist", "christian|disciples_of_christ", "christian|episcopal", "christian|evangelical", "christian|evangelical_covenant", "christian|exclusive_brethren", "christian|lutheran", "christian|mennonite", "christian|methodist", "christian|moravian", "christian|pentecostal", "christian|presbyterian", "christian|quaker", "christian|reformed", "christian|uniting", "christian|polish_national_catholic", "christian|african_methodist_episcopal", "christian|african_methodist_episcopal_zion", "christian|alliance", "christian|apostolic_faith", "christian|assemblies_of_god", "christian|brethren_in_christ", "christian|calvinistic_methodist", "christian|catholic_apostolic", "christian|church_of_god_in_christ", "christian|church_of_scotland", "christian|churches_of_christ", "christian|czechoslovak_hussite", "christian|dutch_reformed", "christian|evangelical_free_church_of_america", "christian|evangelical_lutheran", "christian|evangelical_free_church_of_france", "christian|foursquare", "christian|free_church_of_scotland", "christian|living_waters_church", "christian|mission_covenant_church_of_sweden", "christian|mormon", "christian|nazarene", "christian|new_frontiers", "christian|orthodox_presbyterian_church", "christian|pkn", "christian|remonstrant", "christian|salvation_army", "christian|scottish_episcopal", "christian|seventh_day_adventist", "christian|strict_baptist", "christian|temple_society_australia", "christian|united", "christian|united_free_church_of_scotland", "christian|united_reformed", "christian|united_methodist", "christian|united_church_of_christ", "christian|welsh_baptist", "christian|welsh_independent", "christian|apostolic", "christian|assyrian", "christian|catholic_mariavite", "christian|bosnian_church", "christian|charismatic", "christian|christian_community", "christian|christ_scientist", "christian|church_of_christ", "christian|congregational", "christian|ecumenical", "christian|fsspx", "christian|harrist", "christian|iglesia_ni_cristo", "christian|jehovahs_witness", "christian|kimbanguist", "christian|la_luz_del_mundo", "christian|liberal_catholic", "christian|mariavite", "christian|messianic_jewish", "christian|new_apostolic", "christian|nondenominational", "christian|old_catholic", "christian|philippine_independent", "christian|shizmatic", "christian|simultaneum", "christian|spiritist", # Hindu "hindu|shaktism", "hindu|hare_krishna", "hindu|vaishnavism", "hindu|shaivism", "hindu|smartism", "hindu|ganpatya", "hindu|warkari", "hindu|dattatreya", # Islamic "muslim|ahmadiyya", "muslim|alevi", "muslim|bektashi", "muslim|ibadi", "muslim|ismaili", "muslim|shia", "muslim|sunni", "muslim|sufi", # Jain "jain|digambara", "jain|svetambara", # Jewish "jewish|unaffiliated", "jewish|ashkenazi", "jewish|buchari", "jewish|conservative", "jewish|hasidic", "jewish|kabbalistic", "jewish|karaite", "jewish|lubavitch", "jewish|mizrachi", "jewish|modern_orthodox", "jewish|neo_orthodox", "jewish|orthodox", "jewish|reconstructionist", "jewish|reform", "jewish|progressive", "jewish|liberal", "jewish|samaritan", "jewish|sephardi", "jewish|traditional", "jewish|unity", # Pagan "pagan|asatru", "pagan|baltic", "pagan|celtic", "pagan|greco-roman", "pagan|slavic", "pagan|wicca", # Sikh "sikh|khalsa", "sikh|akali_nihang", "sikh|giani_samparda", "sikh|nirmala_sect", "sikh|sewapanthi", "sikh|udasi", "sikh|ramgarhia", "sikh|radha_soami", "sikh|namdhari", "sikh|nirankari", "sikh|nanakpanthi", "sikh|ramraiyas", # Taoist "taoist|quanzhen", "taoist|zhengyi", # Zoroastrian "zoroastrian|irani", "zoroastrian|parsi", ] ##################################################### # Citation CSS ##################################################### # Yoinked from the OpenWebUI CSS FONTS = ",".join([ "-apple-system", "BlinkMacSystemFont", "Inter", "ui-sans-serif", "system-ui", "Segoe UI", "Roboto", "Ubuntu", "Cantarell", "Noto Sans", "sans-serif", "Helvetica Neue", "Arial", "\"Apple Color Emoji\"", "\"Segoe UI Emoji\"", "Segoe UI Symbol", "\"Noto Color Emoji\"" ]) FONT_CSS = f""" html {{ font-family: {FONTS}; }} @media (prefers-color-scheme: dark) {{ html {{ --tw-text-opacity: 1; color: rgb(227 227 227 / var(--tw-text-opacity)); }} }} """ HIGHLIGHT_CSS = HtmlFormatter().get_style_defs('.highlight') ##################################################### # Useful Constants ##################################################### NOMINATIM_LOOKUP_TYPES = { "node": "N", "route": "R", "way": "W" } OLD_VALVE_SETTING = """ Tell the user that you cannot search OpenStreetMap until the configuration is fixed. The Nominatim URL valve setting needs to be updated. There has been a breaking change in 1.0 of the OpenStreetMap tool. The valve setting is currently set to: `{OLD}`. It shoule be set to the root URL of the Nominatim endpoint, for example: `https://nominatim.openstreetmap.org/` Inform the user they need to fix this configuration setting. """.replace("\n", " ").strip() VALVES_NOT_SET = { "results": [], "instructions": ( "Tell the user that the User-Agent and From headers" "must be set to comply with the OSM Nominatim terms" "of use: https://operations.osmfoundation.org/policies/nominatim/" ).replace("\n", " ").strip() } NO_RESULTS = { "results": [], "instructions": ("No results found. Tell the user you found no results. " "Do not make up answers or hallucinate. Only say you " "found no results.") } NO_RESULTS_BAD_ADDRESS = { "results": [], "instructions": ("No results found. Tell the user you found no results because " "OpenStreetMap could not resolve the address. " "Print the exact address or location you searched for. " "Suggest to the user that they refine their " "question, for example removing the apartment number, sub-unit, " "etc. Example: If `123 Main Street, Apt 4` returned no results, " "suggest that the user searc for `123 Main Street` instead. " "Use the address the user searched for in your example.") } NO_CONFUSION = ("**IMPORTANT!:** Check that the results match the location " "the user is talking about, by analyzing the conversation history. " "Sometimes there are places with the same " "names, but in different cities or countries. If the results are for " "a different city or country than the user is interested in, say so: " "tell the user that the results are for the wrong place, and tell them " "to be more specific in their query.") # Give examples of OSM links to help prevent wonky generated links # with correct GPS coords but incorrect URLs. EXAMPLE_OSM_LINK = "https://www.openstreetmap.org/#map=19/<lat>/<lon>" OSM_LINK_INSTRUCTIONS = ( "When necessary, make friendly human-readable OpenStreetMap links " "by using the latitude and longitude of the amenities: " f"{EXAMPLE_OSM_LINK}\n\n" ) CITATION_INSTRUCTIONS = [ ( "Always use inline citations in the format " "[id], using the citation_id of each source." ), ( "The citation ID must be in the form of [<citation_id>]. " "For example, if the citation_id ID is 123456789, print [123456789]." ), "Do NOT print [id:123456789]. That will not work. ", "Always use inline citations for information relating to a result. " ] ##################################################### # Global Utils ##################################################### def chunk_list(input_list, chunk_size): it = iter(input_list) return list( itertools.zip_longest(*[iter(it)] * chunk_size, fillvalue=None) ) def to_lookup(thing) -> Optional[str]: lookup_type = NOMINATIM_LOOKUP_TYPES.get(thing['type']) if lookup_type is not None: return f"{lookup_type}{thing['id']}" def get_or_none(tags: dict, *keys: str) -> Optional[str]: """ Try to extract a value from a dict by trying keys in order, or return None if none of the keys were found. """ for key in keys: if key in tags: return tags[key] return None def all_are_none(*args) -> bool: for arg in args: if arg is not None: return False return True ##################################################### # Instruction Generation ##################################################### def specific_place_instructions() -> str: return ( "# Result Instructions\n" "These are search results ordered by relevance for the " "address, place, landmark, or location the user is asking " "about. **IMPORTANT!:** Tell the user all relevant information, " "including address, contact information, and the OpenStreetMap link. " "Make the map link into a nice human-readable markdown link." ) def navigation_instructions(travel_type) -> str: return ( "This is the navigation route that the user has requested. " f"These instructions are for travel by {travel_type}. " "Tell the user the total distance, " "and estimated travel time. " "If the user **specifically asked for it**, also tell " "them the route itself. When telling the route, you must tell " f"the user that it's a **{travel_type}** route." ) def list_instructions(tag_type: str, used_rel: bool) -> List[str]: """ Produce detailed instructions in a structured manner for models that support it. """ instructions = { "result_accuracy": [], "information_reporting": [], "links": [], "citations": [] } # basic instructions if used_rel: instructions["result_accuracy"].append( "You **MUST** Begin your reply by explaining to the user " "that we could not search the entire area, " "and therefore we did not get all results." ) instructions["result_accuracy"].append( "Inform the user that more accurate results can be found by " "using more specific search terms like a street or specific landmark." ) else: instructions["result_accuracy"].append("These are the results known to be closest to the requested location.") instructions["result_accuracy"].append("When saying the location name, use the resolved_location field.") # how to report information instructions["information_reporting"].append( "When telling the user about the results, make sure to report " "all information relevant to the user's query (address, contact info, website, etc)." ) instructions["information_reporting"].append( "Do not report information that is irrelvant to the user's query." ) instructions["information_reporting"].append( "Prefer closer results by TRAVEL DISTANCE first. " "Closer results are higher in the list." ) instructions["information_reporting"].append( "When telling the user the distance, use the TRAVEL DISTANCE. Do not say one " "distance is farther away than another. Just say what the distances are. " ) instructions["information_reporting"].append( "Only use relevant results. If there are no relevant results, " "say so. Do not make up answers or hallucinate. " ) instructions["information_reporting"].append(NO_CONFUSION) instructions["information_reporting"].append( "Remember that the CLOSEST result is first, and you should use " "that result first." ) # links and citations instructions["links"].append(OSM_LINK_INSTRUCTIONS) instructions["links"].append( "Give map links friendly, contextual labels. " "Don't just print the naked link. " f"Example: `You can view it on [OpenStreetMap]({EXAMPLE_OSM_LINK})`" ) instructions["citations"].extend(CITATION_INSTRUCTIONS) return instructions def simple_instructions(tag_type_str: str, used_rel: bool) -> str: """ Produce simpler markdown-oriented instructions for models that do better with that. """ if used_rel: rel_inst = ( "**Mention that the central point of the town or city that the user is searching in " "was used, and that results may not cover the whole area.**" ) else: rel_inst = "" return ( f"These are some of the {tag_type_str} points of interest nearby. " "These are the results known to be closest to the requested location. " "For each result, report the following information: \n" " - Name\n" " - Address\n" " - OpenStreetMap Link (make it a human readable link like 'View on OpenStreetMap')\n" " - Contact information (address, phone, website, email, etc)\n\n" "Use the information provided to answer the user's question. " "The results are ordered by closeness as the crow flies. " "When telling the user about distances, use the TRAVEL DISTANCE only. " "Only use relevant results. If there are no relevant results, " "say so. Do not make up answers or hallucinate. " "Make sure that your results are in the actual location the user is talking about, " "and not a place of the same name in a different country." f"{rel_inst}" ) ##################################################### # Event Emission ##################################################### class OsmEventEmitter: def __init__(self, event_emitter, status_indicators=True): self.event_emitter = event_emitter self.status_indicators = status_indicators async def navigating(self, done: bool): if not self.status_indicators: return if done: message = "Navigation complete" else: message = "Navigating..." await self.event_emitter({ "type": "status", "data": { "status": "in_progress", "description": message, "done": done, }, }) async def navigation_error(self, exception: Exception): if not self.status_indicators: return await self.event_emitter({ "type": "status", "data": { "status": "error", "description": f"Error navigating: {str(exception)}", "done": True, }, }) async def resolving(self, done: bool=False, message: Optional[str]=None, items=[]): if not self.status_indicators: return items = [ { "title": get_or_none(item, "display_name"), "link": OsmUtils.create_osm_link(item.get('lat', -1), item.get('lon', -1)) } for item in items ] if done: message = "Resolution complete" else: message = f" location: {message}" if message is not None else "..." message = f"Resolving{message}" await self.event_emitter({ "type": "status", "data": { "action": "web_search", "status": "in_progress", "items": items if items else None, "description": message, "done": done, }, }) async def searching( self, category: str, place: str, status: str="in_progress", done: bool=False, tags=[] ): if not self.status_indicators: return query = f"{category.capitalize()} POIs near {place}" await self.event_emitter({ "type": "status", "data": { "status": status, "action": "web_search_queries_generated", "queries": [query], "description": f"Searching for {category} near {place}", "done": done, }, }) async def search_complete(self, category: str, place: str, items=[]): if not self.status_indicators: return num_results = len(items) citation_items = [] for item in items: name = get_or_none(item.get('tags', {}), "name", "brand") addr = get_or_none(item.get('tags', {}), "addr:street", "addr:city") addr = f"({addr})" if addr else "" title = f"{name} {addr}".strip() osm_link = OsmUtils.create_osm_link(item.get('lat', -1), item.get('lon', -1)) citation_items.append({ "title": title, "link": osm_link }) await self.event_emitter({ "type": "status", "data": { "action": "web_search", "status": "in_progress", "items": citation_items if citation_items else None, "description": f"Found {num_results} results for {category}", "done": True, }, }) await self.event_emitter({ "type": "status", "data": { "action": "sources_retrieved", "status": "in_progress", "count": num_results, "description": f"Found {num_results} results for {category}", "done": True, }, }) async def emit_result_citation(self, thing): if not self.status_indicators: return converted = OsmUtils.create_citation_document(thing) if not converted: return source_name = converted["source_name"] document = converted["document"] osm_link = converted["osm_link"] website = converted["website"] await self.event_emitter({ "type": "source", "data": { "document": [document], "metadata": [{"source": source_name, "html": True }], "source": {"name": website, "url": website}, } }) async def error(self, exception: Exception): if not self.status_indicators: return await self.event_emitter({ "type": "status", "data": { "status": "error", "description": f"Error searching OpenStreetMap: {str(exception)}", "done": True, }, }) ##################################################### # OSM-Related Utility Functions (AKA spaghetti land) ##################################################### class OsmUtils: """ Utility functions that are organized as static methods. AKA dumping ground for spaghetti code. """ @staticmethod def create_citation_document(thing) -> Optional[dict]: if not thing: return None if 'address' in thing: street = get_or_none(thing['address'], "road") else: street = get_or_none(thing['tags'], "addr:street") id = thing.get('id', None) street_name = street if street is not None else "" source_name = f"{thing['name']} {street_name}" lat, lon = thing['lat'], thing['lon'] addr = f"at {thing['address']}" if thing['address'] != 'unknown' else 'nearby' osm_link = OsmUtils.create_osm_link(lat, lon) json_data = OsmUtils.pretty_print_thing_json(thing) website = thing.get('website', osm_link) document = (f"<style>{HIGHLIGHT_CSS}</style>" f"<style>{FONT_CSS}</style>" f"<div>" f"<p>{thing['name']} is located {addr}.</p>" f"<ul>" f"<li>" f" <strong>Opening Hours:</strong> {thing['opening_hours']}" f"</li>" f"</ul>" f"<p>Raw JSON data:</p>" f"{json_data}" f"</div>") return { "id": id, "source_name": source_name, "document": document, "osm_link": osm_link, "website": website } @staticmethod def create_osm_link(lat, lon): return EXAMPLE_OSM_LINK.replace("<lat>", str(lat)).replace("<lon>", str(lon)) @staticmethod def pretty_print_thing_json(thing): """Converts an OSM thing to nice JSON HTML.""" formatted_json_str = json.dumps(thing, indent=2) lexer = JsonLexer() formatter = HtmlFormatter(style='colorful') return highlight(formatted_json_str, lexer, formatter) @staticmethod def thing_is_useful(thing): """ Determine if an OSM way entry is useful to us. This means it has something more than just its main classification tag, and (usually) has at least a name. Some exceptions are made for ways that do not have names. """ tags = thing.get('tags', {}) has_tags = len(tags) > 1 has_useful_tags = ( 'leisure' in tags or 'shop' in tags or 'amenity' in tags or 'car:rental' in tags or 'rental' in tags or 'car_rental' in tags or 'service:bicycle:rental' in tags or 'tourism' in tags ) # there can be a lot of artwork in city centers. drop ones that # aren't as notable. we define notable by the thing having wiki # entries, or by being tagged as historical. if tags.get('tourism', '') == 'artwork': notable = ( 'wikipedia' in tags or 'wikimedia_commons' in tags ) else: notable = True return has_tags and has_useful_tags and notable @staticmethod def thing_has_info(thing): has_name = any('name' in tag for tag in thing['tags']) return OsmUtils.thing_is_useful(thing) and has_name @staticmethod def get_bounding_box_center(bbox): def convert(bbox, key): return bbox[key] if isinstance(bbox[key], float) else float(bbox[key]) min_lat = convert(bbox, 'minlat') min_lon = convert(bbox, 'minlon') max_lat = convert(bbox, 'maxlat') max_lon = convert(bbox, 'maxlon') return { 'lon': (min_lon + max_lon) / 2, 'lat': (min_lat + max_lat) / 2 } @staticmethod def haversine_distance(point1, point2): R = 6371 # Earth radius in kilometers lat1, lon1 = point1['lat'], point1['lon'] lat2, lon2 = point2['lat'], point2['lon'] d_lat = math.radians(lat2 - lat1) d_lon = math.radians(lon2 - lon1) a = (math.sin(d_lat / 2) * math.sin(d_lat / 2) + math.cos(math.radians(lat1)) * math.cos(math.radians(lat2)) * math.sin(d_lon / 2) * math.sin(d_lon / 2)) c = 2 * math.atan2(math.sqrt(a), math.sqrt(1 - a)) return R * c @staticmethod def add_haversine_distance_to_things(origin, things_nearby): for thing in things_nearby: if 'distance' not in thing: thing['distance'] = round(OsmUtils.haversine_distance(origin, thing), 3) ##################################################### # OSM Result Parsing ##################################################### # For each search category, we may want to add additional tags # directly from the raw OSM data for a given search result. This is a # map of search category as defined by the tool functions to OSM tag # names. If a tag is in the list for the category, that tag will be # included in search results' additional_info field (assuming it # exists on the OSM entity). # # Tag values ending with a :* have a special meaning. The tool will # find all "sub-tags" that begin with the prefix. For example, # "socket:*" will find "socket:type2", "socket:type2:output", # "socket:type1" and so on. This can work on "sub-sub-tags" too. # For example, "socket:type2:*" would only look for tags that # start with the prefix "socket:type2". RAW_DATA_BY_CATEGORY = { # stores "groceries": { "organic", "fair_trade", "origin" }, "convenience": { "organic", "fair_trade", "origin", "ice_cream" }, "cannabis": { "cannabis:*" }, "drugs": { "cannabis:*" }, "electronics": { "electronics", "repair", "second_hand" }, # food and drink "sit_down_restaurants": {"cuisine", "reservation"}, "fast_food": {"cuisine"}, "cafe_or_bakery": {"cuisine"}, "bars_and_pubs": {"cocktails", "reservation"}, # recreation "sports": { "sport", "surface", "lit" }, "swimming": { "swimming_lanes", "outdoor", "indoor", "location", "depth", "heated", "lifeguard", "sport:scuba_diving"}, "playgrounds": { "surface", "min_age", "max_age", "fee", "playground=", "playground:*", "stroller", "fenced"}, # tourism "accommodation": { "stars", "internet_access", "rooms", "check_in", "check_out" }, "tourist_attractions": { "museum", "outdoor", "zoo", "fee", "artwork_type", "artist_name" }, # education "schools": { "capacity", "grades", "isced:level", "min_age", "max_age", "religion", "pedagogy", "language", "school:language" }, # healthcare "pharmacies": { "dispensing", "drive_through" }, # fuel etc "fossil_fuels": { "self_service", "full_service", "full_service:conditional", "car_wash", "compressed_air", "toilets", "water_point", "sanitary_dump_station" }, "ev_fast_charging": {"fee", "socket:*", "payment:*", "authentication:*", "motorcar", "truck", "ref" }, } class OsmParser: """All result parsing-related functionality in one place.""" def __init__( self, original_location: str, things_nearby: List[dict], category: Optional[str], include_raw_data=True ): self.original_location = original_location self.things_nearby = things_nearby self.include_raw_data = include_raw_data self.category = category @staticmethod def parse_address_from_address_obj(address) -> Optional[str]: """Parse address from Nominatim address object.""" house_number = get_or_none(address, "house_number") street = get_or_none(address, "road") city = get_or_none(address, "city") state = get_or_none(address, "state") postal_code = get_or_none(address, "postcode") # if all are none, that means we don't know the address at all. if all_are_none(house_number, street, city, state, postal_code): return None # Handle missing values to create complete-ish addresses, even if # we have missing data. We will get either a partly complete # address, or None if all the values are missing. line1 = filter(None, [street, house_number]) line2 = filter(None, [city, state, postal_code]) line1 = " ".join(line1).strip() line2 = " ".join(line2).strip() full_address = filter(None, [line1, line2]) full_address = ", ".join(full_address).strip() return full_address if len(full_address) > 0 else None @staticmethod def parse_address_from_tags(tags: dict) -> Optional[str]: """Parse address from Overpass tags object.""" house_number = get_or_none(tags, "addr:housenumber", "addr:house_number") street = get_or_none(tags, "addr:street") city = get_or_none(tags, "addr:city") state = get_or_none(tags, "addr:state", "addr:province") postal_code = get_or_none( tags, "addr:postcode", "addr:post_code", "addr:postal_code", "addr:zipcode", "addr:zip_code" ) # if all are none, that means we don't know the address at all. if all_are_none(house_number, street, city, state, postal_code): return None # Handle missing values to create complete-ish addresses, even if # we have missing data. We will get either a partly complete # address, or None if all the values are missing. line1 = filter(None, [street, house_number]) line2 = filter(None, [city, state, postal_code]) line1 = " ".join(line1).strip() line2 = " ".join(line2).strip() full_address = filter(None, [line1, line2]) full_address = ", ".join(full_address).strip() return full_address if len(full_address) > 0 else None @staticmethod def parse_thing_address(thing: dict) -> Optional[str]: """ Parse address from either an Overpass result or Nominatim result. """ if 'address' in thing: # nominatim result return OsmParser.parse_address_from_address_obj(thing['address']) else: return OsmParser.parse_address_from_tags(thing['tags']) @staticmethod def friendly_shop_name(shop_type: str) -> str: """ Make certain shop types more friendly for LLM interpretation. """ if shop_type == "doityourself": return "hardware" else: return shop_type @staticmethod def parse_thing_amenity_type(thing: dict, tags: dict) -> Optional[dict]: """ Extract amenity type or other identifying category from Nominatim or Overpass result object. """ if 'amenity' in tags: return tags['amenity'] if thing.get('class') == 'amenity' or thing.get('class') == 'shop': return thing.get('type') # fall back to tag categories, like shop=* if 'shop' in tags: return OsmParser.friendly_shop_name(tags['shop']) if 'leisure' in tags: return OsmParser.friendly_shop_name(tags['leisure']) return None @staticmethod def extract_raw_data(category: str, thing: dict) -> dict: tags_to_extract = RAW_DATA_BY_CATEGORY.get(category, None) additional_info = {} if not tags_to_extract: return additional_info # loop through all tags in thing. if regular tag to extract, # add to dict. if wildcard, do a beginswith check. additiona_info = {} tags: dict = thing.get('tags', {}) for tag_name in tags_to_extract: if tag_name.endswith(':*'): tag_prefix = tag_name.split(':*')[0] + ":" for tag, value in tags.items(): if tag.startswith(tag_prefix): additional_info[tag] = value else: for tag, value in tags.items(): if tag == tag_name: additional_info[tag] = value return additional_info @staticmethod def parse_and_validate_thing(thing: dict, category: Optional[str]) -> Optional[dict]: """ Parse an OSM result (node or post-processed way) and make it more friendly to work with. Helps remove ambiguity of the LLM interpreting the raw JSON data. If there is not enough data, discard the result. """ tags: dict = thing.get('tags', {}) # Currently we define "enough data" as at least having lat, lon, # and a name. nameless things are allowed if they are in a certain # class of POIs (leisure). has_name = 'name' in tags or 'name' in thing is_leisure = 'leisure' in tags or 'leisure' in thing if 'lat' not in thing or 'lon' not in thing: return None if not has_name and not is_leisure: return None lat: float = thing['lat'] lon: float = thing['lon'] friendly_thing = {} name: str = (tags['name'] if 'name' in tags else thing['name'] if 'name' in thing else str(thing['id']) if 'id' in thing else str(thing['osm_id']) if 'osm_id' in thing else "unknown") osm_link = OsmUtils.create_osm_link(lat, lon) amenity_type: Optional[str] = OsmParser.parse_thing_amenity_type(thing, tags) address: str = OsmParser.parse_thing_address(thing) distance: Optional[float] = thing.get('distance') nav_distance: Optional[float] = thing.get('nav_distance') opening_hours: Optional[str] = tags.get('opening_hours') operator: Optional[str] = tags.get('operator') access: Optional[str] = tags.get('access') brand: Optional[str] = tags.get('brand') website: Optional[str] = get_or_none( tags, "operator:website", "website", "brand:website" ) # use the navigation distance if it's present. but if not, set to # the haversine distance so that we at least get coherent results # for LLM. friendly_thing['distance'] = "{:.3f}".format(distance) if distance else "unknown" if nav_distance: friendly_thing['nav_distance'] = "{:.3f}".format(nav_distance) + " km" else: friendly_thing['nav_distance'] = f"a bit more than {friendly_thing['distance']}km" if brand: friendly_thing['brand'] = brand if operator: friendly_thing['operator'] = operator if access: friendly_thing['access'] = access friendly_thing['osm_id'] = thing.get('id', 'unknown') friendly_thing['osm_type'] = thing.get('type', None) friendly_thing['name'] = name if name else "unknown" friendly_thing['address'] = address if address else "unknown" friendly_thing['lat'] = lat if lat else "unknown" friendly_thing['lon'] = lon if lon else "unknown" friendly_thing['amenity_type'] = amenity_type if amenity_type else "unknown" friendly_thing['opening_hours'] = opening_hours if opening_hours else "not recorded" friendly_thing['website'] = website if website else osm_link friendly_thing['phone_number'] = thing.get('phone', 'unknown') if category: friendly_thing['additional_info'] = OsmParser.extract_raw_data(category, thing) return friendly_thing def convert_and_validate_results( self, sort_message: str="closeness", use_distance: bool=True ) -> Optional[str]: """ Converts search results into something friendlier for LLM understanding. Also drops incomplete results. Supports Overpass and Nominatim results. """ original_location = self.original_location things_nearby = self.things_nearby entries = [] for thing in things_nearby: # Convert to friendlier data, drop results without names etc. # No need to instruct LLM to generate map links if we do it # instead. friendly_thing = OsmParser.parse_and_validate_thing(thing, self.category) if not friendly_thing: continue map_link = OsmUtils.create_osm_link(friendly_thing['lat'], friendly_thing['lon']) hv_distance_json = (f"{friendly_thing['distance']} km" if use_distance else "unavailable") trv_distance_json = (f"{friendly_thing['nav_distance']}" if use_distance and 'nav_distance' in friendly_thing else "unavailable") # prepare friendly thing for final result entry. some # values are on the "raw" JSON from earlier in the search # process. thing.pop('nav_distance', None) thing.pop('distance', None) citation_id = thing.pop('citation_id', None) friendly_thing.update({ "geographical_distance": hv_distance_json, "travel_distance": trv_distance_json, "openstreetmap_link": map_link, "citation_id": citation_id, }) if self.include_raw_data: friendly_thing["raw_osm_json"] = thing entries.append(friendly_thing) if len(entries) == 0: return None return entries ##################################################### # Caching of OSM Data ##################################################### class OsmCache: def __init__(self, filename="/tmp/osm.json"): self.filename = filename self.data = {} # Load existing cache if it exists try: with open(self.filename, 'r') as f: self.data = json.load(f) except FileNotFoundError: pass def get(self, key): return self.data.get(key) def set(self, key, value): self.data[key] = value with open(self.filename, 'w') as f: json.dump(self.data, f) def get_or_set(self, key, func_to_call): """ Retrieve the value from the cache for a given key. If the key is not found, call `func_to_call` to generate the value and store it in the cache. :param key: The key to look up or set in the cache :param func_to_call: A callable function that returns the value if key is missing :return: The cached or generated value """ if key not in self.data: value = func_to_call() self.set(key, value) return self.data[key] def clear_cache(self): """ Clear all entries from the cache. """ self.data.clear() try: # Erase contents of the cache file. with open(self.filename, 'w'): pass except FileNotFoundError: pass ##################################################### # Navigation Routing ##################################################### class OrsRouter: def __init__( self, valves, user_valves, events: OsmEventEmitter ): self.cache = OsmCache() self.events = events if valves.ors_api_key is not None and valves.ors_api_key != "": if valves.ors_instance is not None: self._client = openrouteservice.Client( base_url=valves.ors_instance, key=valves.ors_api_key ) else: self._client = openrouteservice.Client(key=valves.ors_api_key) else: self._client = None def calculate_route( self, from_thing: dict, to_thing: dict ) -> Optional[dict]: """ Calculate route between A and B. Returns the route, if successful, or None if the distance could not be calculated, or if ORS is not configured. """ if not self._client: return None # select profile based on distance for more accurate # measurements. very close haversine distances use the walking # profile, which should (usually?) essentially cover walking # and biking. further away = use car. if to_thing.get('distance', 9000) <= 1.5: profile = "foot-walking" else: profile = "driving-car" coords = ((from_thing['lon'], from_thing['lat']), (to_thing['lon'], to_thing['lat'])) # check cache first. cache_key = f"ors_route_{str(coords)}" cached_route = self.cache.get(cache_key) if cached_route: print("[OSM] Got route from cache!") return cached_route resp = ors_directions(self._client, coords, profile=profile, preference="fastest", units="km") routes = resp.get('routes', []) if len(routes) > 0: self.cache.set(cache_key, routes[0]) return routes[0] else: return None def calculate_distance( self, from_thing: dict, to_thing: dict ) -> Optional[float]: """ Calculate navigation distance between A and B. Returns the distance calculated, if successful, or None if the distance could not be calculated, or if ORS is not configured. """ if not self._client: return None route = self.calculate_route(from_thing, to_thing) return route.get('summary', {}).get('distance', None) if route else None ##################################################### # Main Search Logic ##################################################### class OsmSearcher: def __init__(self, valves, user_valves, event_emitter): self.events = event_emitter # config settings self.nominatim_url = valves.nominatim_url self.overpass_turbo_url = valves.overpass_turbo_url self.user_agent = valves.user_agent self.from_header = valves.from_header self.include_raw_data = valves.include_raw_osm_data self.detailed_instructions = (user_valves.instruction_oriented_interpretation if user_valves else valves.instruction_oriented_interpretation) # dependents self._ors = OrsRouter(valves, user_valves, event_emitter) @staticmethod def sort_by_rank(things): """ Calculate a rank for a list of things. More important ones are pushed towards the top. Currently only for tourism tags. """ def rank_thing(thing: dict) -> int: tags = thing.get('tags', {}) if not 'tourism' in tags: return 0 rank = len([name for name in tags.keys() if name.startswith("name")]) rank += 5 if 'historic' in tags else 0 rank += 5 if 'wikipedia' in tags else 0 rank += 1 if 'wikimedia_commons' in tags else 0 rank += 5 if tags.get('tourism', '') == 'museum' else 0 rank += 5 if tags.get('tourism', '') == 'aquarium' else 0 rank += 5 if tags.get('tourism', '') == 'zoo' else 0 return rank return sorted(things, reverse=True, key=lambda t: (rank_thing(t), -t['distance'])) @staticmethod def sort_by_closeness(origin, points, *keys: str): """ Sorts a list of { lat, lon }-like dicts by closeness to an origin point. The origin is a dict with keys of { lat, lon }. """ return sorted(points, key=itemgetter(*keys)) @staticmethod def process_way_result(way) -> Optional[dict]: """ Post-process an OSM Way dict to remove the geometry and node info, and calculate a single GPS coordinate from its bounding box. """ if 'nodes' in way: del way['nodes'] if 'geometry' in way: del way['geometry'] if 'bounds' in way: way_center = OsmUtils.get_bounding_box_center(way['bounds']) way['lat'] = way_center['lat'] way['lon'] = way_center['lon'] del way['bounds'] return way return None @staticmethod def flatten_tags(tag_query): """ Make sure that the tags queried results in a completely flat list of OSM tags. That is, remove AND groups. """ flat_tags = [] for str_or_list in tag_query: if isinstance(str_or_list, str): flat_tags.append(str_or_list) else: flat_tags += [tag for tag in str_or_list] return flat_tags @staticmethod def group_tags(tags): """ De-duplicate tags of the same type into a single Overpass Turbo search term. Also pluck out anything that needs to be ANDed together. """ result = {} and_queries = [] for tag in tags: # skip AND queries because they must be handled # separately. if not isinstance(tag, str): and_queries.append(tag) continue key, value = tag.split('=') if key not in result: result[key] = [] result[key].append(value) return result, and_queries @staticmethod def fallback(nominatim_result): """ If we do not have Overpass Turbo results, attempt to use the Nominatim result instead. """ return ([nominatim_result] if 'type' in nominatim_result and (nominatim_result['type'] == 'amenity' or nominatim_result['type'] == 'shop' or nominatim_result['type'] == 'leisure' or nominatim_result['type'] == 'tourism') else []) @staticmethod def merge_from_nominatim(thing, nominatim_result) -> Optional[dict]: """Merge information into object missing all or some of it.""" if thing is None: return None if 'address' not in nominatim_result: return None nominatim_address = nominatim_result['address'] # prioritize actual name, road name, then display name. display # name is often the full address, which is a bit much. nominatim_name = nominatim_result.get('name') nominatim_road = nominatim_address.get('road') nominatim_display_name = nominatim_result.get('display_name') thing_name = thing.get('name') if nominatim_name and not thing_name: thing['name'] = nominatim_name.strip() elif nominatim_road and not thing_name: thing['name'] = nominatim_road.strip() elif nominatim_display_name and not thing_name: thing['name'] = nominatim_display_name.strip() tags = thing.get('tags', {}) for key in nominatim_address: obj_key = f"addr:{key}" if obj_key not in tags: tags[obj_key] = nominatim_address[key] thing['tags'] = tags return thing def create_headers(self) -> Optional[dict]: if len(self.user_agent) == 0 or len(self.from_header) == 0: return None return { 'User-Agent': self.user_agent, 'From': self.from_header } def calculate_navigation_distance(self, start, destination) -> float: """Calculate real distance from A to B, instead of Haversine.""" return self._ors.calculate_distance(start, destination) def add_nav_distance_to_things(self, origin, things_nearby: List[dict]) -> bool: """Update distances to use ORS navigable distances, if ORS enabled.""" used_ors = False cache = OsmCache() for thing in things_nearby: cache_key = f"ors_{origin}_{thing['id']}" nav_distance = cache.get(cache_key) if nav_distance: print(f"[OSM] Got nav distance for {thing['id']} from cache!") else: print(f"[OSM] Checking ORS for {thing['id']}") try: nav_distance = self.calculate_navigation_distance(origin, thing) except Exception as e: print(f"[OSM] Error querying ORS: {e}") print("[OSM] Falling back to regular distance due to ORS error!") nav_distance = thing['distance'] if nav_distance: used_ors = True cache.set(cache_key, nav_distance) thing['nav_distance'] = round(nav_distance, 3) return used_ors def get_result_instructions(self, tag_type_str: str, used_rel: bool) -> str: if self.detailed_instructions: return list_instructions(tag_type_str, used_rel) else: return simple_instructions(tag_type_str, used_rel) def calculate_bounding_box(self, nominatim_result): """ Determine bounding box and/or point to use for starting search. """ rel_types = ["village", "town", "city", "suburb"] # relations can have a lat/lon defined on them. useful for or # other settlements. only do this if the nominatim result is a # relation that defines an entire settled area. # see also OSM admin_level rel_osm_type = nominatim_result.get("osm_type", None) == "relation" rel_type = nominatim_result.get("type", None) == "administrative" rel_class = nominatim_result.get("class", None) == "boundary" rel_addr_type = nominatim_result.get("addresstype", None) in rel_types use_rel_bbox = rel_osm_type and rel_type and rel_class and rel_addr_type if use_rel_bbox: rel_lat = nominatim_result.get("lat", None) rel_lon = nominatim_result.get("lon", None) if rel_lat and rel_lon: print("[OSM] Requested search area is settled urban zone. Using lat/lon of OSM relation.") return True, { 'minlat': rel_lat, 'maxlat': rel_lat, 'minlon': rel_lon, 'maxlon': rel_lon } # fall back to nominatim bounding box for everything else. return False, { 'minlat': nominatim_result['boundingbox'][0], 'maxlat': nominatim_result['boundingbox'][1], 'minlon': nominatim_result['boundingbox'][2], 'maxlon': nominatim_result['boundingbox'][3] } async def enrich_from_nominatim(self, things, format="json"): """ Enrich search results missing data with extra info from Nominatim. """ updated_things = [] # the things with merged info. # handle last chunk, which can have nones due to the way # chunking is done. things = [thing for thing in things if thing is not None] lookups = [] for thing in things: if thing is None: continue lookup = to_lookup(thing) if lookup is not None: lookups.append(lookup) # Nominatim likes it if we cache our data. cache = OsmCache() lookups_to_remove = [] for lookup_id in lookups: from_cache = cache.get(lookup_id) if from_cache is not None: updated_things.append(from_cache) lookups_to_remove.append(lookup_id) # only need to look up things we do not have cached. lookups = [id for id in lookups if id not in lookups_to_remove] if len(lookups) == 0: print("[OSM] Got all Nominatim info from cache!") return updated_things else: print(f"[OSM] Looking up {len(lookups)} things from Nominatim") url = urljoin(self.nominatim_url, "lookup") params = { 'osm_ids': ",".join(lookups), 'format': format } headers = self.create_headers() if not headers: raise ValueError("Headers not set") response = requests.get(url, params=params, headers=headers) if response.status_code == 200: data = response.json() if not data: print("[OSM] No results found for lookup") return [] addresses_by_id = {item['osm_id']: item for item in data} for thing in things: nominatim_result = addresses_by_id.get(thing['id'], {}) if nominatim_result != {}: updated = OsmSearcher.merge_from_nominatim(thing, nominatim_result) if updated is not None: lookup = to_lookup(thing) cache.set(lookup, updated) updated_things.append(updated) return updated_things else: await self.events.error(Exception(response.text)) print(response.text) return [] async def nominatim_search(self, query, format="json", limit: int=1) -> Optional[dict]: await self.events.resolving(done=False, message=query) cache_key = f"nominatim_search_{query}" cache = OsmCache() data = cache.get(cache_key) if data: print(f"[OSM] Got nominatim search data for {query} from cache!") await self.events.resolving(done=True, message=query, items=data[:limit]) return data[:limit] print(f"[OSM] Searching Nominatim for: {query}") url = urljoin(self.nominatim_url, "search") params = { 'q': query, 'format': format, 'addressdetails': 1, 'limit': limit, } headers = self.create_headers() if not headers: await self.events.error("Headers not set") raise ValueError("Headers not set") response = requests.get(url, params=params, headers=headers) if response.status_code == 200: data = response.json() if not data: raise ValueError(f"No results found for query '{query}'") await self.events.resolving(done=True, message=query, items=data[:limit]) print(f"[OSM] Got result from Nominatim for: {query}") cache.set(cache_key, data[:limit]) return data[:limit] else: await self.events.error(Exception(response.text)) print(response.text) return None async def search_osm( self, place, tags, bbox, limit=5, radius=4000, not_tag_groups={} ) -> (List[dict], List[dict]): """ Return a list relevant of OSM nodes and ways. Some post-processing is done on ways in order to add coordinates to them. Optionally specify not_tag_groups, which is a dict of tag to values, all of which will be excluded from the search. Each item in the tag list is treated as an OR query. An entry in the tag list can be a string (basic search for "this tag"), or it can be a list of strings itself, which is treated as an AND query. """ print(f"[OSM] Searching Overpass Turbo around origin {place}") headers = self.create_headers() if not headers: raise ValueError("Headers not set") center = OsmUtils.get_bounding_box_center(bbox) around = f"(around:{radius},{center['lat']},{center['lon']})" tag_groups, and_queries = OsmSearcher.group_tags(tags) # tags is e.g. [amenity=restaurant, place=blah, ...]. # and_queries is a list of those lists. These tags will be # ANDed in Overpass, i.e. the search must fulfill all # conditions. Normally, we OR instead. and_groups = ["".join(f"[{tag}]" for tag in tags) for tags in and_queries] search_groups = [f'"{tag_type}"~"{"|".join(values)}"' for tag_type, values in tag_groups.items()] not_search_groups = [f'"{tag_type}"!~"{"|".join(values)}"' for tag_type, values in not_tag_groups.items()] not_search_query = "".join([f"[{query}]" for query in not_search_groups]) searches = [] for search_group in search_groups: searches.append(f'nwr[{search_group}]{not_search_query}{around}') for and_group in and_groups: searches.append(f'nwr{and_group}{not_search_query}{around}') search = ";\n".join(searches) if len(search) > 0: search += ";" # "out geom;" is needed to get bounding box info of ways, # so we can calculate the coordinates. query = f""" [out:json]; ( {search} ); out geom; """ print(query) data = { "data": query } response = requests.get(self.overpass_turbo_url, params=data, headers=headers) if response.status_code == 200: # nodes have have exact GPS coordinates. we also include # useful way entries, post-processed to remove extra data # and add a centered calculation of their GPS coords. any # way that doesn't have enough info for us to use is # dropped. results = response.json() results = results['elements'] if 'elements' in results else [] nodes = [] ways = [] things_missing_names = [] for res in results: if 'type' not in res or not OsmUtils.thing_is_useful(res): continue if res['type'] == 'node': if OsmUtils.thing_has_info(res): nodes.append(res) else: things_missing_names.append(res) elif res['type'] == 'way': processed = OsmSearcher.process_way_result(res) if processed is not None and OsmUtils.thing_has_info(res): ways.append(processed) else: if processed is not None: things_missing_names.append(processed) # attempt to update ways that have no names/addresses. if len(things_missing_names) > 0: print(f"[OSM] Updating {len(things_missing_names)} things with info") for way_chunk in chunk_list(things_missing_names, 20): updated = await self.enrich_from_nominatim(way_chunk) ways = ways + updated return nodes, ways else: print(response.text) raise Exception(f"Error calling Overpass API: {response.text}") async def search_and_rank( self, nominatim_result, place, tags, bbox, limit, radius, not_tag_groups): """ Search OpenStreetMap and sort results according to distance and relevance to the query. """ # initial OSM search nodes, ways = await self.search_osm( place, tags, bbox, limit, radius, not_tag_groups ) # use results from the initial search, but if they do not # exist, fall back to the nominatim result. this may or may # not be a good idea. things_nearby = (nodes + ways if len(nodes) > 0 or len(ways) > 0 else OsmSearcher.fallback(nominatim_result)) # add distance to things and drop number of results to the # limit. then, if enabled, we calculate ORS distances. then we # sort again. origin = OsmUtils.get_bounding_box_center(bbox) OsmUtils.add_haversine_distance_to_things(origin, things_nearby) # sort by importance + distance, drop to the requested limit, # then sort by closeness. things_nearby = OsmSearcher.sort_by_rank(things_nearby) things_nearby = things_nearby[:limit] # drop down to requested limit things_nearby = OsmSearcher.sort_by_closeness(origin, things_nearby, 'distance') if self.add_nav_distance_to_things(origin, things_nearby): sort_method = "travel distance" things_nearby = OsmSearcher.sort_by_closeness(origin, things_nearby, 'nav_distance', 'distance') else: sort_method = "haversine distance" # post-process to add sequential citation/source IDs, because # web ui doesn't like arbitrary source IDs. for id, thing in enumerate(things_nearby): thing["citation_id"] = str(id + 1) return [things_nearby, sort_method] async def search_nearby( self, place: str, tags: List[str], category: str, limit: int=5, radius: int=4000, not_tag_groups={}, readable_category: str = "POIs" ) -> dict: """ Main entrypoint into the searching logic. Checks header validity, and then uses Nominatim to figure out WHERE to search. From that spot, search for POIs using Overpass Turbo (with data enrichment from Nominatim if necessary), rank them by distance and relevance, and then final result list plus some metadata. """ headers = self.create_headers() if not headers: return { "place_display_name": place, "results": VALVES_NOT_SET } try: nominatim_result = await self.nominatim_search(place, limit=1) except ValueError: nominatim_result = [] if not nominatim_result or len(nominatim_result) == 0: await self.events.search_complete(readable_category, place, []) return { "place_display_name": place, "results": NO_RESULTS_BAD_ADDRESS } try: nominatim_result = nominatim_result[0] # display friendlier searching message if possible if 'display_name' in nominatim_result: place_display_name = ",".join(nominatim_result['display_name'].split(",")[:3]) elif 'address' in nominatim_result: addr = OsmParser.parse_thing_address(nominatim_result) if addr is not None: place_display_name = ",".join(addr.split(",")[:3]) else: place_display_name = place else: print(f"[OSM] WARN: Could not find display name for place: {place}") place_display_name = place await self.events.searching(readable_category, place_display_name, done=False, tags=tags) used_rel, bbox = self.calculate_bounding_box(nominatim_result) print(f"[OSM] Searching for {readable_category} near {place_display_name}") things_nearby, sort_method = await self.search_and_rank( nominatim_result, place, tags, bbox, limit, radius, not_tag_groups ) if not things_nearby or len(things_nearby) == 0: await self.events.search_complete(readable_category, place_display_name, []) return { "place_display_name": place, "results": NO_RESULTS } print(f"[OSM] Found {len(things_nearby)} {readable_category} results near {place_display_name}") tag_type_str = ", ".join(OsmSearcher.flatten_tags(tags)) # Only print the full result instructions if we # actually have something. parser = OsmParser(place, things_nearby, category=category, include_raw_data=self.include_raw_data) search_results = parser.convert_and_validate_results(sort_message=sort_method) if search_results: result_instructions = self.get_result_instructions(tag_type_str, used_rel) else: result_instructions = "No results found at all. Tell the user there are no results." resp = { "instructions": result_instructions, "results": search_results if search_results else [] } # emit citations for the actual results. await self.events.search_complete(readable_category, place_display_name, things_nearby) for thing in search_results: await self.events.emit_result_citation(thing) return { "place_display_name": place_display_name, "results": resp, "things": things_nearby } except ValueError: await self.events.search_complete(readable_category, place_display_name, []) return { "place_display_name": place_display_name, "results": NO_RESULTS, "things": [] } except Exception as e: print(e) await self.events.error(e) instructions = (f"No results were found, because of an error. " f"Tell the user that there was an error finding results. " f"**Do not make up results. There are no results available because of an error.**") result = { "instructions": instructions, "results": [], "error_message": f"{e}" } return { "place_display_name": place_display_name, "results": result, "things": [] } async def do_osm_search( valves, user_valves, place, tags, category="POIs", event_emitter=None, limit=5, radius=4000, setting='urban', not_tag_groups={}, radius_multiplier=1, ): # handle breaking 1.0 change, in case of old Nominatim valve settings. if valves.nominatim_url.endswith("/search"): message = "Old Nominatim URL setting still in use!" print(f"[OSM] ERROR: {message}") if valves.status_indicators and event_emitter is not None: await event_emitter({ "type": "status", "data": { "status": "error", "description": f"Error searching OpenStreetMap: {message}", "done": True, }, }) return { "instructions": OLD_VALVE_SETTING.replace("{OLD}", valves.nominatim_url), "results": [] } print(f"[OSM] Searching for [{category}] ({tags[0]}, etc) near place: {place} ({setting} setting)") radius = radius * setting_to_multiplier(setting) * radius_multiplier events = OsmEventEmitter(event_emitter) readable_category=category.replace("_", " ") searcher = OsmSearcher(valves, user_valves, events) search = await searcher.search_nearby(place=place, tags=tags, limit=limit, radius=radius, category=category, not_tag_groups=not_tag_groups, readable_category=readable_category) # move place display name to top level results returned to client. place_display_name = search.get("place_display_name", None) search.get("results", {}).update(resolved_location=place_display_name) return search["results"] ##################################################### # Navigation Logic ##################################################### class OsmNavigator: def __init__( self, valves, user_valves: Optional[dict], events: OsmEventEmitter, ): self.valves = valves self.user_valves = user_valves self.events = events async def navigate(self, start_place: str, destination_place: str): await self.events.navigating(done=False) searcher = OsmSearcher(self.valves, self.user_valves, self.events) router = OrsRouter(self.valves, self.user_valves, self.events) try: start = await searcher.nominatim_search(start_place, limit=1) destination = await searcher.nominatim_search(destination_place, limit=1) if not start or not destination: await self.events.navigating(done=True) return NO_RESULTS start, destination = start[0], destination[0] route = router.calculate_route(start, destination) if not route: await self.evnts.event_navigating(done=True) return NO_RESULTS total_distance = round(route.get('summary', {}).get('distance', ''), 2) travel_time = round(route.get('summary', {}).get('duration', 0) / 60.0, 2) travel_type = "car" if total_distance > 1.5 else "walking/biking" def create_step_instruction(step): instruction = step['instruction'] duration = round(step.get('duration', 0.0) / 60.0, 2) distance = round(step.get('distance', 0.0), 2) if duration <= 0.0 or distance <= 0.0: return f"- {instruction}" if duration < 1.0: duration = f"{round(duration * 60.0, 2)} sec" else: duration = f"{duration} min" if distance < 1.0: distance = f"{round(distance * 1000.0, 2)}m" else: distance = f"{distance}km" return f"- {instruction} ({distance}, {duration})" nav_instructions = [ create_step_instruction(step) for segment in route["segments"] for step in segment["steps"] ] result = { "instructions": navigation_instructions(travel_type), "travel_method": travel_type, "total_distance": f"{total_distance} km", "travel_time": f"{str(travel_time)} minutes", "route": nav_instructions } await self.events.navigating(done=True) return result except Exception as e: print(e) await self.events.navigation_error(e) message = (f"There are no results due to an error. " "Tell the user that there was an error.") return { "instructions": message, "error_message": f"{e}", "travel_method": None, "total_distance": None, "travel_time": None, "route": None } ##################################################### # Normalization & Validation Functions ##################################################### def normalize_setting(setting: Optional[str]) -> str: if setting: setting = setting.lower().replace('"', '').replace("'", '') if setting in ['urban', 'suburban', 'rural']: return setting else: print(f"[OSM] WARN: {setting} is not a valid urban setting. Defaulting to 'urban'.") return 'urban' return 'urban' def setting_to_multiplier(setting: str) -> int: if setting == 'urban': return 1 elif setting == 'suburban': return 2 elif setting == 'rural': return 3.5 return 1 # The various *_to_tags functions convert the LLM-requested thing to # search for into a set of OSM tags. Most of the time, these are just # simple lists of tags, where any OSM entity having at least one tag # is returned. But the functions can also return ANDed tags (entities # must have ALL specified tags), and tags that are rejected outright # (entities must NOT have those tags). def store_category_to_tags(store_type: str) -> Tuple[List[str], dict]: if store_type == "groceries": return ["shop=supermarket", "shop=grocery", "shop=greengrocer"], {} elif store_type == "convenience": return ["shop=convenience"], {} elif store_type == "alcohol": return ["shop=alcohol"], {} elif store_type == "drugs" or store_type == "cannabis": return ["shop=coffeeshop", "shop=cannabis", "shop=headshop", "shop=smartshop"], {} elif store_type == "electronics": return ["shop=electronics", "shop=computer", "shop=hifi", "shop=mobile_phone", "shop=radiotechnics"], {} elif store_type == "electrical": return ["shop=lighting", "shop=electrical"], {} elif store_type == "hardware" or store_type == "diy": return ["shop=doityourself", "shop=hardware", "shop=power_tools", "shop=groundskeeping", "shop=trade"], {} elif store_type == "pharmacies": return ["amenity=pharmacy", "shop=chemist", "shop=supplements", "shop=health_food"], {} else: return [], {} def recreation_to_tags(recreation_type: str) -> Tuple[List[str], dict]: if recreation_type == "swimming": return ["leisure=swimming_pool", "leisure=swimming_area", "leisure=water_park", "tourism=theme_park"], {} elif recreation_type == "playgrounds": return ["leisure=playground"], {} elif recreation_type == "amusement": return ["leisure=park", "leisure=amusement_arcade", "tourism=theme_park"], {} elif recreation_type == "gyms": # gyms can either show up as leisure=fitness_centre, or a # combo of leisure=sports_centre and sport=fitness. return ["leisure=fitness_centre", "leisure=fitness_center", ["leisure=sports_centre", "sport=fitness" ]], {} elif recreation_type == "sports": # gyms are often tag with sports_center + leisure fitness. but # we want to search for those separately. return ["leisure=horse_riding", "leisure=ice_rink", "leisure=disc_golf_course", "leisure=pitch", "leisure=sports_center", "leisure=sports_centre", "leisure=sports_hall"], { "sport": [ "fitness" ] } else: return [], {} def food_category_to_tags(food_type: str) -> Tuple[List[str], dict]: if food_type == "sit_down_restaurants": return ["amenity=restaurant", "amenity=eatery", "amenity=canteen"], {} elif food_type == "fast_food": return ["amenity=fast_food"], {} elif food_type == "cafe_or_bakery": # cannabis stores are sometimes tagged with amenity=cafe. do # not return them when searching for food to eat. return ["shop=bakery", "amenity=cafe"], { "shop": [ "cannabis" ] } elif food_type == "bars_and_pubs": # the search for amenity=pub can also wind up finding # public_bookcases. we don't want that. return ["amenity=bar", "amenity=pub", "amenity=biergarten"], { "amenity": [ "public_bookcase" ] } else: return [], {} def travel_category_to_tags(travel_type: str) -> Tuple[List[str], dict]: if travel_type == "tourist_attractions": return ["tourism=museum", "tourism=aquarium", "tourism=zoo", "tourism=attraction", "tourism=gallery", "tourism=artwork"], {} elif travel_type == "accommodation": return ["tourism=hotel", "tourism=chalet", "tourism=guest_house", "tourism=guesthouse", "tourism=motel", "tourism=hostel"], {} elif travel_type == "bike_rentals": return ["amenity=bicycle_rental", "amenity=bicycle_library", "service:bicycle:rental=yes"], {} elif travel_type == "car_rentals": return ["amenity=car_rental", "car:rental=yes", "rental=car", "car_rental=yes"], {} elif travel_type == "public_transport": return ["highway=bus_stop", "amenity=bus_station", "railway=station", "railway=halt", "railway=tram_stop", "station=subway", "amenity=ferry_terminal", "public_transport=station"], {} else: return [], {} def healthcare_category_to_tags(healthcare_type: str) -> Tuple[List[str], dict]: if healthcare_type == "doctor": return ["amenity=clinic", "amenity=doctors", "healthcare=doctor"], {} elif healthcare_type == "dentist": return ["amenity=dentist", "healthcare=dentist", "healthcare:speciality=orthodontics"], {} elif healthcare_type == "hospital": return ["healthcare=hospital", "amenity=hospitals"], {} elif healthcare_type == "pharmacy": return ["amenity=pharmacy", "shop=chemist"], {} else: return [] def education_category_to_tags(education_type: str) -> Tuple[List[str], dict]: if education_type == "schools": return ["amenity=school"], {} elif education_type == "universities_and_colleges": return ["amenity=university", "amenity=college"], {} elif education_type == "libraries": return ["amenity=library"], {} else: return [], {} def validate_category(category, allowed_categories): if category not in allowed_categories: return { "results": [], "instructions": "There was an error. Attempt to correct the error, or inform the user (whichever is appropriate).", "error_message": f"{category} is not a valid category. Must be one of: {', '.join(allowed_categories)}" } return None def validate_tags(category, tags, allowed_categories): category_validation_error = validate_category(category, allowed_categories) if category_validation_error: return category_validation_error if not tags: return { "results": [], "instructions": "There was an error. Attempt to correct the error, or inform the user (whichever is appropriate).", "error_message": f"{category} translated to no searchable OSM tags." } return None # EV charger mappings, moved out of original function for readability. EV_CHARGER_TOOL_INPUT_CHARGER_TYPES = { "chademo": "chademo", "chademo3": "chaoji", "chaoji": "chaoji", "ccs2": "type2_combo", "ccs1": "type1_combo", "nacs": "nacs", "gb/t": "gb_dc" } # socket types used on OSM. EV_CHARGER_OSM_SOCKET_TYPES = [ # Various DC fast charging standards "chademo", # chademo "chaoji", # chademo3 "type2_combo", # CCS2 "type1_combo", # CCS1 "gb_dc", # chinese standard "nacs", # north american standard # not supposed to be used, but can show up. "tesla_supercharger", "tesla_supercharger_ccs", "ccs" ] ##################################################### # Main Tool Entrypoint ##################################################### # # This is the part we've all been waiting for... # class Tools: class Valves(BaseModel): user_agent: str = Field( default="", description="Unique user agent to identify your OSM API requests." ) from_header: str = Field( default="", description="Email address to identify your OSM requests." ) nominatim_url: str = Field( default="https://nominatim.openstreetmap.org/", description="URL of OSM Nominatim API for reverse geocoding (address lookup)." ) overpass_turbo_url: str = Field( default="https://overpass-api.de/api/interpreter", description="URL of Overpass Turbo API for searching OpenStreetMap." ) instruction_oriented_interpretation: bool = Field( default=True, description=("Give detailed result interpretation instructions to the model. " "Switch this off if results are inconsistent, wrong, or missing.") ) ors_api_key: Optional[str] = Field( default=None, description=("Provide an Open Route Service API key to calculate " "more accurate distances (leave default to disable).") ) ors_instance: Optional[str] = Field( default=None, description="Use a custom ORS instance (leave default to use public ORS instance)." ) status_indicators: bool = Field( default=True, description=("Emit status update events to the web UI.") ) include_raw_osm_data: bool = Field( default=False, description="Include raw OSM data in search results or not." ) tile_server: str = Field( default="https://{s}.tile.openstreetmap.org/{z}/{x}/{y}.png", description="OSM Tile Server" ) pass class UserValves(BaseModel): instruction_oriented_interpretation: bool = Field( default=True, description=("Give detailed result interpretation instructions to the model. " "Switch this off if results are inconsistent, wrong, or missing.") ) pass def __init__(self): self.valves = self.Valves() self.user_valves = None async def find_address_for_coordinates(self, latitude: float, longitude: float, __event_emitter__) -> str: """ Resolves GPS coordinates to a specific address or place. :param latitude: The latitude portion of the GPS coordinate. :param longitude: The longitude portion of the GPS coordinate. :return: Information about the address or place at the coordinates. """ print(f"[OSM] Resolving [{latitude}, {longitude}] to address.") return await self.find_specific_place(f"{latitude}, {longitude}", __event_emitter__) async def find_specific_named_business_or_landmark_near_coordinates( self, store_or_business_name: str, latitude: float, longitude: float, __event_emitter__ ) -> str: """ Finds specifically named stores, businesses, or landmarks near the given GPS coordinates. This can be used for general or specific things, like 'gas station' or 'grocery store', or the name of a specific chain (like 'Wal-Mart' or 'Albert Heijn'). This function is intended for finding specific chains or businesses near the given coordinates. Use this if the user asks about businesses or places nearby. :param store_or_business_name: Name of store or business to look for. :param latitude: The latitude portion of the GPS coordinate. :param longitude: The longitude portion of the GPS coordinate. :return: Information about the address or places near the coordinates. """ print(f"[OSM] Searching for '{store_or_business_name}' near {latitude},{longitude}") query = f"{store_or_business_name} {latitude},{longitude}" return await self.find_specific_place(query, __event_emitter__) async def resolve_coordinates(self, latitude: float, longitude: float, __event_emitter__) -> str: """ Resolves GPS coordinates to a human-readable location such as an address or place name. Use this to convert GPS coordinates before searching if you need to figure out urban, suburban, or rural setting. :param latitude: The latitude portion of the GPS coordinate. :param longitude: The longitude portion of the GPS coordinate. :return: Reverse geocoded location or address information. """ print(f"Reverse geocoding {latitude},{longitude}") return await self.find_specific_place(f"{latitude},{longitude}", __event_emitter__) async def find_specific_place(self, address_or_place: str, __event_emitter__) -> str: """ Looks up details on OpenStreetMap of a specific address, landmark, place, named building, or location. Used for when the user asks where a specific unique entity (like a specific museum, or church, or shopping center) is. :param address_or_place: The address or place to look up. :return: Information about the place, if found. """ print(f"[OSM] Searching for info on [{address_or_place}].") events = OsmEventEmitter(__event_emitter__) searcher = OsmSearcher(self.valves, self.user_valves, events) try: result = await searcher.nominatim_search(address_or_place, limit=5) if result: parser = OsmParser(address_or_place, result, category=None) results = parser.convert_and_validate_results( sort_message="importance", use_distance=False ) resp = { "instructions": specific_place_instructions(), "results": results if results else [] } return resp else: return NO_RESULTS except Exception as e: print(e) return (f"There are no results due to an error. " "Tell the user that there was an error. " f"The error was: {e}. " f"Tell the user the error message.") async def navigate_between_places( self, start_address_or_place: str, destination_address_or_place: str, __event_emitter__ ) -> str: """ Retrieve a navigation route and associated information between two places. :param start_address_or_place: The address, place, or coordinates to start from. :param destination_address_or_place: The destination address, place, or coordinates to go. :return: The navigation route and associated info, if found. """ print(f"[OSM] Navigating from [{start_address_or_place}] to [{destination_address_or_place}].") events = OsmEventEmitter(__event_emitter__) navigator = OsmNavigator(self.valves, self.user_valves, events) return await navigator.navigate(start_address_or_place, destination_address_or_place) async def find_stores_by_category_near_place( self, place: str, category: StoreCategory, setting: UrbanSetting, __user__: dict, __event_emitter__ ) -> str: """ Finds stores of a specific type on OpenStreetMap near a given place or address. For setting, specify if the place is an urban area, a suburb, or a rural location. :param place: The name of a place, an address, or GPS coordinates. City and country must be specified, if known. :param setting: Urban-ness of the requested location. Controls search radius. :param category: Category of store to search for. """ allowed_categories = [ "groceries", "convenience", "alcohol", "drugs", "cannabis", "electronics", "electrical", "hardware", "diy" ] setting = normalize_setting(setting) user_valves = __user__["valves"] if "valves" in __user__ else None tags, not_tag_groups = store_category_to_tags(category) validation_error = validate_tags(category, tags, allowed_categories) if validation_error: return validation_error return await do_osm_search(valves=self.valves, user_valves=user_valves, category=category, setting=setting, place=place, tags=tags, event_emitter=__event_emitter__, not_tag_groups=not_tag_groups) async def find_recreation_by_category_near_place( self, place: str, category: RecreationCategory, setting: UrbanSetting, __user__: dict, __event_emitter__ ) -> str: """ Finds recreation facilities of a specific type on OpenStreetMap near a given place or address. For setting, specify if the place is an urban area, a suburb, or a rural location. Note: amusement category can be arcades, theme parks, and similar. :param place: The name of a place, an address, or GPS coordinates. City and country must be specified, if known. :param setting: Urban-ness of the requested location. Controls search radius. :param category: Category of recreation to search for. """ allowed_categories = ["swimming", "playgrounds", "amusement", "sports", "gyms"] setting = normalize_setting(setting) user_valves = __user__["valves"] if "valves" in __user__ else None tags, not_tag_groups = recreation_to_tags(category) validation_error = validate_tags(category, tags, allowed_categories) if validation_error: return validation_error radius = 4000 limit = 5 if category == "swimming": radius = 10000 elif category == "amusement" or category == "sports": radius = 10000 limit = 10 elif category == "playgrounds": limit = 10 return await do_osm_search(valves=self.valves, user_valves=user_valves, category=category, radius=radius, limit=limit, setting=setting, place=place, tags=tags, event_emitter=__event_emitter__, not_tag_groups=not_tag_groups) async def find_eateries_and_drinking_by_category_near_place( self, place: str, category: EateryCategory, setting: UrbanSetting, __user__: dict, __event_emitter__ ) -> str: """ Finds places to eat or drink on OpenStreetMap near a given place or address. For setting, specify if the place is an urban area, a suburb, or a rural location. If it is unclear what category the user wants, ask for clarification. :param place: The name of a place, an address, or GPS coordinates. City and country must be specified, if known. :param setting: Urban-ness of the requested location. Controls search radius. :param category: Category of eateries to search for. """ allowed_categories = [ "sit_down_restaurants", "fast_food", "cafe_or_bakery", "bars_and_pubs"] setting = normalize_setting(setting) user_valves = __user__["valves"] if "valves" in __user__ else None tags, not_tag_groups = food_category_to_tags(category) validation_error = validate_tags(category, tags, allowed_categories) if validation_error: return validation_error return await do_osm_search(valves=self.valves, user_valves=user_valves, category=category, limit=10, setting=setting, place=place, tags=tags, event_emitter=__event_emitter__, not_tag_groups=not_tag_groups) async def find_travel_info_by_category_near_place( self, place: str, category: TravelCategory, setting: UrbanSetting, __user__: dict, __event_emitter__ ) -> str: """ Find tourist attractions, accommodation, public transport, bike rentals, or car rentals on OpenStreetMap. For setting, specify if the place is an urban area, a suburb, or a rural location. If it is unclear what category the user wants, ask for clarification. :param place: The name of a place, an address, or GPS coordinates. City and country must be specified, if known. :param setting: Urban-ness of the requested location. Controls search radius. :param category: Category of travel info to search for. """ allowed_categories = ["tourist_attractions", "accommodation", "bike_rentals", "car_rentals", "public_transport"] setting = normalize_setting(setting) user_valves = __user__["valves"] if "valves" in __user__ else None tags, not_tag_groups = travel_category_to_tags(category) validation_error = validate_tags(category, tags, allowed_categories) if validation_error: return validation_error radius = 4000 limit = 5 if category == "car_rentals": radius = 6000 elif category == "tourist_attractions": radius = 10000 limit = 10 elif category == "accommodation": radius = 10000 elif category == "public_transport": limit = 10 return await do_osm_search(valves=self.valves, user_valves=user_valves, category=category, limit=limit, radius=radius, setting=setting, place=place, tags=tags, event_emitter=__event_emitter__, not_tag_groups=not_tag_groups) async def find_healthcare_by_category_near_place( self, place: str, category: HealthcareCategory, setting: UrbanSetting, __user__: dict, __event_emitter__ ) -> str: """ Find healthcare, doctors, dentists, hospitals, and pharmacies on OpenStreetMap. For setting, specify if the place is an urban area, a suburb, or a rural location. If it is unclear what category the user wants, ask for clarification. :param place: The name of a place, an address, or GPS coordinates. City and country must be specified, if known. :param setting: Urban-ness of the requested location. Controls search radius. :param category: Category of healthcare to search for. """ allowed_categories = ["doctor", "dentist", "hospital", "pharmacy"] setting = normalize_setting(setting) user_valves = __user__["valves"] if "valves" in __user__ else None tags, not_tag_groups = healthcare_category_to_tags(category) validation_error = validate_tags(category, tags, allowed_categories) if validation_error: return validation_error return await do_osm_search(valves=self.valves, user_valves=user_valves, category=category, setting=setting, place=place, tags=tags, event_emitter=__event_emitter__, not_tag_groups=not_tag_groups) async def find_place_of_worship_near_place( self, __user__: dict, place: str, religion: ReligionCategory, denomination: DenominationCategory, is_common_for_area: bool, setting: UrbanSetting, __event_emitter__ ) -> str: """ Finds places of worship (churches, mosques, temples, etc) on OpenStreetMap near a given place or address. For setting, specify if the place is an urban area, a suburb, or a rural location. If searching for a denomination, the denomination must be part of the given religion. If NOT searching for any specific denomination, use the special 'any_denomination' value. If a religion is specified as not common in the requested area, the search radius will be greatly increased. A religion is common in the requested area if it traditionally has a lot of adherents or a visible social presence in the area. :param place: The name of a place, an address, or GPS coordinates. City and country must be specified, if known. :param religion: Narrow the search to a particular religion. Use "any_religion" to search for all places of worship. :param denomination: Narrow search to denomination of particular religion. Denomination must match religion. :param is_common_for_area: Whether or not the religion or denomination is common in the requested area. :param setting: Urban-ness of the requested location. Controls search radius. :return: A list of nearby places of worship, if found. """ setting = normalize_setting(setting) user_valves = __user__["valves"] if "valves" in __user__ else None # by default, look for all places of worship. tags = ["amenity=place_of_worship"] # searching for a specific religion? if religion != "any_religion" and religion is not None: tags = [["amenity=place_of_worship", f"religion={religion}"]] # or even a specific religion + denomination? if denomination != 'any_denomination' and denomination is not None: split = denomination.split("|") if len(split) != 2 or '|' not in denomination: raise ValueError("Denomination must be specified as 'religion|denomination'") denom_religion = split[0] search_by_denom = split[1] if denom_religion != religion: raise ValueError(f'Denomination specified. religion parameter must be: {denom_religion}') # now we can actually search by religion + denom tags = [["amenity=place_of_worship", f"religion={religion}", f"denomination={search_by_denom}"]] radius_multiplier = 1 if is_common_for_area else 2.5 return await do_osm_search(valves=self.valves, user_valves=user_valves, category="places of worship", setting=setting, place=place, tags=tags, event_emitter=__event_emitter__, radius_multiplier=radius_multiplier) async def find_education_by_category_near_place( self, place: str, category: EducationCategory, setting: UrbanSetting, __user__: dict, __event_emitter__ ) -> str: """ Find education institutions and resources on OpenStreetMap. For setting, specify if the place is an urban area, a suburb, or a rural location. Note: category "schools" searches only for primary and secondary schools, NOT universities or colleges. :param place: The name of a place, an address, or GPS coordinates. City and country must be specified, if known. :param setting: Urban-ness of the requested location. Controls search radius. :param category: Category of education to search for. """ allowed_categories = ["schools", "universities_and_colleges", "libraries"] setting = normalize_setting(setting) user_valves = __user__["valves"] if "valves" in __user__ else None tags, not_tag_groups = education_category_to_tags(category) validation_error = validate_tags(category, tags, allowed_categories) if validation_error: return validation_error return await do_osm_search(valves=self.valves, user_valves=user_valves, category=category, setting=setting, place=place, tags=tags, event_emitter=__event_emitter__, not_tag_groups=not_tag_groups) async def find_fuel_or_charging_by_category_near_place( self, place: str, category: FuelCategory, fuel_type: FuelTypeCategory, ev_charger_type: EVChargerCategory, setting: UrbanSetting, __user__: dict, __event_emitter__ ) -> str: """ Finds gas stations, petrol stations, fuel stations, or EV fast chargers near a given place or address. For setting, specify if the place is an urban area, a suburb, or a rural location. Does not find slow (regular) EV chargers. EV fast charger note: 'chademo3' is not the same as 'chademo'. 'chademo3' is also known as 'chaoji'.j :param place: The name of a place, an address, or GPS coordinates. City and country must be specified, if known. :param setting: Urban-ness of the requested location. Controls search radius. :param category: Category to search for: either fossil fuels or EV fast chargers. :param fuel_type: Type of fossil fuel to search for. :param ev_charger_type: Type of EV charger to search for. :return: A list of nearby fueling stations, if found. """ # subfunction to find fuel async def _find_fossil_fuel_near_place(valves, __user__: dict, place: str, setting: str, __event_emitter__) -> str: setting = normalize_setting(setting) user_valves = __user__["valves"] if "valves" in __user__ else None tags = ["amenity=fuel"] return await do_osm_search(valves=valves, user_valves=user_valves, category="fossil_fuels", setting=setting, radius=10000, place=place, tags=tags, event_emitter=__event_emitter__) # subfunction to find EV fast chargers. async def _find_ev_fast_chargers_near_place_with_type( valves, __user__: dict, place: str, charger_type: str, setting: str, __event_emitter__ ) -> str: setting = normalize_setting(setting) user_valves = __user__["valves"] if "valves" in __user__ else None charger_type = charger_type.lower().replace('"', "").replace("'", "") # normally, search for any of the possible chargers. tags = [f"socket:{charger}=\\.*" for charger in EV_CHARGER_OSM_SOCKET_TYPES] # or, constrain search? if charger_type != "all": if charger_type not in EV_CHARGER_TOOL_INPUT_CHARGER_TYPES.keys(): return { "results": [], "instructions": "There was an error searching due to an invalid charger type.", "error_message": f"{charger_type} is not a valid charger type." } osm_socket_type = EV_CHARGER_TOOL_INPUT_CHARGER_TYPES[charger_type] tags = [ f"socket:{osm_socket_type}=\\.*" ] return await do_osm_search(valves=valves, user_valves=user_valves, category="ev_fast_charging", setting=setting, radius=10000, place=place, tags=tags, event_emitter=__event_emitter__) # Main part of fuel search function allowed_categories = ["fossil_fuels", "ev_fast_charging"] validation_error = validate_category(category, allowed_categories) if validation_error: return validation_error if category == "fossil_fuels": print(f"[OSM] WARN: Currently ignoring fuel type parameter (was '{fuel_type}')") return await _find_fossil_fuel_near_place(self.valves, __user__, place, setting, __event_emitter__) elif category == "ev_fast_charging": return await _find_ev_fast_chargers_near_place_with_type( self.valves, __user__, place, ev_charger_type, setting, __event_emitter__ ) def show_map(self, category: POICategory, results: List[OsmPOI]) -> HTMLResponse: """ Shows a map for search results. Takes a list of Points of Interest that were returned in a previous search function call. You **must** call another function first to get information required for this function! Make sure to pass all necessary results into this function! When specifying descriptions, make sure they are short and useful, mentioning unique or important info about each result. **After calling this function, only acknowledge that the map is generated. It will be rendered by the client!** :param category: Results category searched for. :param results: Point of Interest (POI) results from a previous function call. :return: HTML for displaying a search results map. """ headers = {"Content-Disposition": "inline"} results_json = json.dumps(results) ui_html = (MAP_UI .replace("{REPLACE_WITH_POI_RESULTS}", results_json) .replace("{REPLACE_WITH_CATEGORY}", category) .replace("{REPLACE_WITH_TILE_SERVER}", self.valves.tile_server)) return HTMLResponse(content=ui_html, headers=headers) # This function exists to help catch situations where the user is # too generic in their query, or is looking for something the tool # does not yet support. By having the model pick this function, we # can direct it to report its capabilities and tell the user how # to use it. It's not perfect, but it works sometimes. def find_other_things_near_place( self, __user__: dict, place: str, setting: UrbanSetting, category: str ) -> str: """ Find shops and other places not covered by a specific category available in the other functions. Use this if the user is asking for a type of store or place that other functions do not support. For setting, specify if the place is an urban area, a suburb, or a rural location. :param place: The name of a place, an address, or GPS coordinates. City and country must be specified, if known. :param setting: Urban-ness of the requested location. Controls search radius. :param category: The category of place, shop, etc to look up. :return: A list of nearby shops. """ print(f"[OSM] Generic catch handler called with {category}") resp = ( "# No Results Found\n" f"No results were found. There was an error. Finding {category} points of interest " "is not yet supported. Tell the user support will come eventually! " "Tell the user that you are only capable of finding specific " "categories of stores, amenities, and points of interest:\n" " - Car rentals and bike rentals\n" " - Public transport, libraries\n" " - Education institutions (schools and universities)\n" " - Grocery stores, supermarkets, convenience stores\n" " - Food and restaurants\n" " - Accommodation\n" " - Places of Worship\n" " - Hardware stores and home improvement centers\n" " - Electrical and lighting stores\n" " - Consumer electronics stores\n" " - Healthcare (doctors, hospitals, pharmacies, and health stores)\n" " - Fuel & EV charging (gas stations, DC fast chargers)\n" " - Various recreational and leisure activities\n\n" "Only mention things from the above list that you think the user " "will be interested in, given the conversation so far. Don't mention " "things not on the list. " "**IMPORTANT**: Tell the user to be specific in their " "query in their next message, so you can call the right function!") return resp