nyc311-describegpt.json

{
  "Dictionary": {
    "response": {
      "fields": [
        {
          "name": "Unique Key",
          "type": "Integer",
          "label": "Record ID",
          "description": "A surrogate primary key – a unique integer that identifies each complaint record.",
          "min": "11465364",
          "max": "48478173",
          "cardinality": 1000000,
          "enumeration": null,
          "null_count": 0,
          "content_type": "unique_id",
          "examples": "<ALL_UNIQUE>"
        },
        {
          "name": "Created Date",
          "type": "DateTime",
          "label": "Complaint Created Date",
          "description": "The calendar date when the complaint was first entered into the system.",
          "min": "01/01/2010",
          "max": "12/23/2020",
          "cardinality": 841014,
          "enumeration": null,
          "null_count": 0,
          "content_type": "date:%m/%d/%Y",
          "examples": "Other… [997333]\n01/24/2013 [347]\n01/07/2014 [315]\n01/08/2015 [283]\n02/16/2015 [269]"
        },
        {
          "name": "Closed Date",
          "type": "DateTime",
          "label": "Complaint Closed Date",
          "description": "The calendar date on which the complaint case was resolved or closed by the agency.",
          "min": "01/01/1900",
          "max": "01/01/2100",
          "cardinality": 688837,
          "enumeration": null,
          "null_count": 28619,
          "content_type": "date:%m/%d/%Y",
          "examples": "Other… [968671]\n(NULL)… [28619]\n11/15/2010 [384]\n11/07/2012 [329]\n12/09/2010 [267]"
        },
        {
          "name": "Agency",
          "type": "String",
          "label": "Handling Agency (abbreviation)",
          "description": "Two‑ to four‑letter code identifying the agency that processed the complaint (e.g., NYPD, DOT).",
          "min": "3-1-1",
          "max": "TLC",
          "cardinality": 28,
          "enumeration": null,
          "null_count": 0,
          "content_type": "category",
          "examples": "NYPD [265116]\nHPD [258033]\nDOT [132462]\nDSNY [81606]\nDEP [75895]"
        },
        {
          "name": "Agency Name",
          "type": "String",
          "label": "Handling Agency (full name)",
          "description": "The full official name of the agency responsible for the complaint.",
          "min": "3-1-1",
          "max": "Valuation Policy",
          "cardinality": 553,
          "enumeration": null,
          "null_count": 0,
          "content_type": "company_name",
          "examples": "New York City Police Depa… [265038]\nDepartment of Housing Pre… [258019]\nDepartment of Transportat… [132462]\nOther… [103974]\nDepartment of Environment… [75895]"
        },
        {
          "name": "Complaint Type",
          "type": "String",
          "label": "Primary Complaint Category",
          "description": "Broad category describing the nature of the complaint, such as noise or parking issues.",
          "min": "../../WEB-INF/web.xml;x=",
          "max": "ZTESTINT",
          "cardinality": 287,
          "enumeration": null,
          "null_count": 0,
          "content_type": "category",
          "examples": "Other… [563561]\nNoise - Residential [89439]\nHEAT/HOT WATER [56639]\nIllegal Parking [45032]\nBlocked Driveway [42356]"
        },
        {
          "name": "Descriptor",
          "type": "String",
          "label": "Complaint Descriptor",
          "description": "A more specific label that refines the primary complaint type (e.g., \"Loud Music/Party\" under Noise).",
          "min": "1 Missed Collection",
          "max": "unknown odor/taste in drinking water (QA6)",
          "cardinality": 1392,
          "enumeration": null,
          "null_count": 3001,
          "content_type": "category",
          "examples": "Other… [671870]\nLoud Music/Party [93646]\nENTIRE BUILDING [36885]\nHEAT [35088]\nNo Access [31631]"
        },
        {
          "name": "Location Type",
          "type": "String",
          "label": "Physical Location Category",
          "description": "Classification of the place where the incident occurred (e.g., residential building, street).",
          "min": "1-, 2- and 3- Family Home",
          "max": "Wooded Area",
          "cardinality": 162,
          "enumeration": null,
          "null_count": 239131,
          "content_type": "category",
          "examples": "RESIDENTIAL BUILDING [255562]\n(NULL)… [239131]\nStreet/Sidewalk [145653]\nResidential Building/Hous… [92765]\nStreet [92190]"
        },
        {
          "name": "Incident Zip",
          "type": "String",
          "label": "Incident ZIP Code",
          "description": "Five‑digit ZIP code for the area where the complaint was reported.",
          "min": "*",
          "max": "XXXXX",
          "cardinality": 535,
          "enumeration": null,
          "null_count": 54978,
          "content_type": "zip_code",
          "examples": "Other… [815988]\n(NULL)… [54978]\n11226 [17114]\n10467 [14495]\n11207 [12872]"
        },
        {
          "name": "Incident Address",
          "type": "String",
          "label": "Full Incident Address",
          "description": "Street address or intersection describing the exact location of the incident. Includes street number, name and directional prefixes when available.",
          "min": "* *",
          "max": "west 155 street and edgecombe avenue",
          "cardinality": 341996,
          "enumeration": null,
          "null_count": 174700,
          "content_type": "street_address",
          "examples": "Other… [819046]\n(NULL)… [174700]\n655 EAST  230 STREET [1538]\n78-15 PARSONS BOULEVARD [694]\n672 EAST  231 STREET [663]"
        },
        {
          "name": "Street Name",
          "type": "String",
          "label": "Primary Street Name",
          "description": "The main street involved in the incident; a component of the full Incident Address.",
          "min": "*",
          "max": "wyckoff avenue",
          "cardinality": 14837,
          "enumeration": null,
          "null_count": 174720,
          "content_type": "street_name",
          "examples": "Other… [784684]\n(NULL)… [174720]\nBROADWAY [9702]\nGRAND CONCOURSE [5851]\nOCEAN AVENUE [3946]"
        },
        {
          "name": "Cross Street 1",
          "type": "String",
          "label": "First Cross Street",
          "description": "First cross‑street intersecting at the incident location; part of the address context.",
          "min": "1 AVE",
          "max": "mermaid",
          "cardinality": 16238,
          "enumeration": null,
          "null_count": 320401,
          "content_type": "street_name",
          "examples": "Other… [619743]\n(NULL)… [320401]\nBEND [12562]\nBROADWAY [8548]\n3 AVENUE [6154]"
        },
        {
          "name": "Cross Street 2",
          "type": "String",
          "label": "Second Cross Street",
          "description": "Second cross‑street intersecting at the incident location, if applicable.",
          "min": "1 AVE",
          "max": "surf",
          "cardinality": 16486,
          "enumeration": null,
          "null_count": 323644,
          "content_type": "street_name",
          "examples": "Other… [623363]\n(NULL)… [323644]\nBEND [12390]\nBROADWAY [8833]\nDEAD END [5626]"
        },
        {
          "name": "Intersection Street 1",
          "type": "String",
          "label": "First Intersection Street",
          "description": "One of the streets that form an intersection at the incident site; used when Address Type = INTERSECTION.",
          "min": "1 AVE",
          "max": "flatlands AVE",
          "cardinality": 11237,
          "enumeration": null,
          "null_count": 767422,
          "content_type": "street_name",
          "examples": "(NULL)… [767422]\nOther… [214544]\nBROADWAY [3761]\nCARPENTER AVENUE [2918]\nBEND [2009]"
        },
        {
          "name": "Intersection Street 2",
          "type": "String",
          "label": "Second Intersection Street",
          "description": "The other street forming the intersection at the incident site.",
          "min": "1 AVE",
          "max": "glenwood RD",
          "cardinality": 11674,
          "enumeration": null,
          "null_count": 767709,
          "content_type": "street_name",
          "examples": "(NULL)… [767709]\nOther… [215667]\nBROADWAY [3462]\nBEND [1942]\n2 AVENUE [1690]"
        },
        {
          "name": "Address Type",
          "type": "String",
          "label": "Address Entry Type",
          "description": "Code indicating how the address was entered (e.g., ADDRESS, INTERSECTION, LATLONG).",
          "min": "ADDRESS",
          "max": "PLACENAME",
          "cardinality": 6,
          "enumeration": null,
          "null_count": 125802,
          "content_type": "category",
          "examples": "ADDRESS [710380]\nINTERSECTION [133361]\n(NULL)… [125802]\nBLOCKFACE [22620]\nLATLONG [7421]"
        },
        {
          "name": "City",
          "type": "String",
          "label": "Incident City / Borough",
          "description": "Name of the city or borough where the incident took place.",
          "min": "*",
          "max": "YORKTOWN HEIGHTS",
          "cardinality": 382,
          "enumeration": null,
          "null_count": 61963,
          "content_type": "city",
          "examples": "BROOKLYN [296254]\nNEW YORK [189069]\nBRONX [181168]\nOther… [163936]\n(NULL)… [61963]"
        },
        {
          "name": "Landmark",
          "type": "String",
          "label": "Nearby Landmark",
          "description": "Notable landmark or feature close to the incident location; may be a public building, park, or other point of interest.",
          "min": "1 AVENUE",
          "max": "ZULETTE AVENUE",
          "cardinality": 5915,
          "enumeration": null,
          "null_count": 912779,
          "content_type": "free_text",
          "examples": "(NULL)… [912779]\nOther… [80165]\nEAST  230 STREET [1545]\nEAST  231 STREET [1291]\nBROADWAY [1148]"
        },
        {
          "name": "Facility Type",
          "type": "String",
          "label": "Related Facility Category",
          "description": "Type of facility involved in the complaint (e.g., garage, school district).",
          "min": "DSNY Garage",
          "max": "School District",
          "cardinality": 6,
          "enumeration": null,
          "null_count": 145478,
          "content_type": "category",
          "examples": "N/A [628279]\nPrecinct [193259]\n(NULL)… [145478]\nDSNY Garage [32310]\nSchool [617]"
        },
        {
          "name": "Status",
          "type": "String",
          "label": "Complaint Status",
          "description": "Current state of the complaint workflow (e.g., Open, Pending, Closed).",
          "min": "Assigned",
          "max": "Unspecified",
          "cardinality": 10,
          "enumeration": "Assigned\nClosed\nClosed - Testing\nEmail Sent\nIn Progress\nOpen\nPending\nStarted\nUnassigned\nUnspecified",
          "null_count": 0,
          "content_type": "category",
          "examples": "Closed [952522]\nPending [20119]\nOpen [12340]\nIn Progress [7841]\nAssigned [6651]"
        },
        {
          "name": "Due Date",
          "type": "DateTime",
          "label": "Resolution Deadline",
          "description": "Date and time by which the agency is expected to resolve or complete action on the complaint.",
          "min": "01/02/1900 12:00:00 AM",
          "max": "06/17/2021 04:34:13 PM",
          "cardinality": 345077,
          "enumeration": null,
          "null_count": 647794,
          "content_type": "datetime:%m/%d/%Y %I:%M:%S %p",
          "examples": "(NULL)… [647794]\nOther… [350746]\n04/08/2015 10:00:58 AM [214]\n05/02/2014 03:32:17 PM [183]\n03/30/2018 10:10:39 AM [172]"
        },
        {
          "name": "Resolution Description",
          "type": "String",
          "label": "Resolution Narrative",
          "description": "Free‑text summary of actions taken, outcomes, or follow‑up instructions for the complaint.",
          "min": "A DOB violation was issued for failing to comply with an existing Stop Work Order.",
          "max": "Your request was submitted to the Department of Homeless Services. The City?s outreach team will assess the homeless individual and offer appropriate assistance within 2 hours. If you asked to know the outcome of your request, you will get a call within 2 hours. No further status will be available through the NYC 311 App, 311, or 311 Online.",
          "cardinality": 1216,
          "enumeration": null,
          "null_count": 20480,
          "content_type": "free_text",
          "examples": "Other… [511739]\nThe Police Department res… [91408]\nThe Department of Housing… [72962]\nThe Police Department res… [63868]\nService Request status fo… [52155]"
        },
        {
          "name": "Resolution Action Updated Date",
          "type": "DateTime",
          "label": "Last Resolution Update Date",
          "description": "Date on which the most recent resolution action was recorded.",
          "min": "12/31/2009",
          "max": "12/23/2020",
          "cardinality": 690314,
          "enumeration": null,
          "null_count": 15072,
          "content_type": "date:%m/%d/%Y",
          "examples": "Other… [982148]\n(NULL)… [15072]\n11/15/2010 [385]\n11/07/2012 [336]\n12/09/2010 [273]"
        },
        {
          "name": "Community Board",
          "type": "String",
          "label": "Community Board Identifier",
          "description": "Code and name of the community board that has jurisdiction over the incident area.",
          "min": "0 Unspecified",
          "max": "Unspecified STATEN ISLAND",
          "cardinality": 77,
          "enumeration": null,
          "null_count": 0,
          "content_type": "category",
          "examples": "Other… [751635]\n0 Unspecified [49878]\n12 MANHATTAN [29845]\n12 QUEENS [23570]\n01 BROOKLYN [21714]"
        },
        {
          "name": "BBL",
          "type": "String",
          "label": "Borough‑Block‑Lot (BBL)",
          "description": "Ten‑digit code uniquely identifying a NYC property parcel; used for spatial referencing.",
          "min": "0000000000",
          "max": "5080470043",
          "cardinality": 268383,
          "enumeration": null,
          "null_count": 243046,
          "content_type": "unknown",
          "examples": "Other… [750668]\n(NULL)… [243046]\n2048330028 [1566]\n4068290001 [696]\n4015110001 [664]"
        },
        {
          "name": "Borough",
          "type": "String",
          "label": "Incident Borough",
          "description": "Name of the borough where the incident occurred.",
          "min": "BRONX",
          "max": "Unspecified",
          "cardinality": 6,
          "enumeration": "BRONX\nBROOKLYN\nMANHATTAN\nQUEENS\nSTATEN ISLAND\nUnspecified",
          "null_count": 0,
          "content_type": "city",
          "examples": "BROOKLYN [296081]\nQUEENS [228818]\nMANHATTAN [195488]\nBRONX [180142]\nUnspecified [49878]"
        },
        {
          "name": "X Coordinate (State Plane)",
          "type": "Integer",
          "label": "State‑Plane X Coordinate",
          "description": "Horizontal coordinate in the New York State Plane system for the incident location.",
          "min": "913281",
          "max": "1067220",
          "cardinality": 102556,
          "enumeration": null,
          "null_count": 85327,
          "content_type": "unknown",
          "examples": "Other… [908535]\n(NULL)… [85327]\n1022911 [1568]\n1037000 [701]\n1023174 [675]"
        },
        {
          "name": "Y Coordinate (State Plane)",
          "type": "Integer",
          "label": "State‑Plane Y Coordinate",
          "description": "Vertical coordinate in the New York State Plane system for the incident location.",
          "min": "121152",
          "max": "271876",
          "cardinality": 116092,
          "enumeration": null,
          "null_count": 85327,
          "content_type": "unknown",
          "examples": "Other… [908538]\n(NULL)… [85327]\n264242 [1566]\n202363 [706]\n211606 [665]"
        },
        {
          "name": "Open Data Channel Type",
          "type": "String",
          "label": "Submission Channel",
          "description": "Medium through which the complaint was submitted (e.g., PHONE, ONLINE, MOBILE).",
          "min": "MOBILE",
          "max": "UNKNOWN",
          "cardinality": 5,
          "enumeration": "MOBILE\nONLINE\nOTHER\nPHONE\nUNKNOWN",
          "null_count": 0,
          "content_type": "category",
          "examples": "PHONE [497606]\nUNKNOWN [230402]\nONLINE [177334]\nMOBILE [79892]\nOTHER [14766]"
        },
        {
          "name": "Park Facility Name",
          "type": "String",
          "label": "Associated Park Facility",
          "description": "Name of a NYC park facility linked to the incident, if applicable.",
          "min": "\"Uncle\" Vito F. Maranzano Glendale Playground",
          "max": "Zimmerman Playground",
          "cardinality": 1889,
          "enumeration": null,
          "null_count": 0,
          "content_type": "category",
          "examples": "Unspecified [993141]\nOther… [5964]\nCentral Park [261]\nRiverside Park [136]\nProspect Park [129]"
        },
        {
          "name": "Park Borough",
          "type": "String",
          "label": "Park Borough",
          "description": "Borough where the referenced park facility is located.",
          "min": "BRONX",
          "max": "Unspecified",
          "cardinality": 6,
          "enumeration": "BRONX\nBROOKLYN\nMANHATTAN\nQUEENS\nSTATEN ISLAND\nUnspecified",
          "null_count": 0,
          "content_type": "city",
          "examples": "BROOKLYN [296081]\nQUEENS [228818]\nMANHATTAN [195488]\nBRONX [180142]\nUnspecified [49878]"
        },
        {
          "name": "Vehicle Type",
          "type": "String",
          "label": "Vehicle Category Involved",
          "description": "Type of vehicle at the incident site (e.g., Car Service, Taxi).",
          "min": "Ambulette / Paratransit",
          "max": "Green Taxi",
          "cardinality": 5,
          "enumeration": null,
          "null_count": 999652,
          "content_type": "category",
          "examples": "(NULL)… [999652]\nCar Service [317]\nAmbulette / Paratransit [19]\nCommuter Van [11]\nGreen Taxi [1]"
        },
        {
          "name": "Taxi Company Borough",
          "type": "String",
          "label": "Taxi Company Operating Borough",
          "description": "Borough where the taxi company operates.",
          "min": "BRONX",
          "max": "Staten Island",
          "cardinality": 11,
          "enumeration": null,
          "null_count": 999156,
          "content_type": "city",
          "examples": "(NULL)… [999156]\nBROOKLYN [207]\nQUEENS [194]\nMANHATTAN [171]\nBRONX [127]"
        },
        {
          "name": "Taxi Pick Up Location",
          "type": "String",
          "label": "Taxi Pickup Point",
          "description": "Location from which a taxi was picked up for the incident; may be an address, intersection, or airport.",
          "min": "1 5 AVENUE MANHATTAN",
          "max": "YORK AVENUE AND EAST 70 STREET",
          "cardinality": 1903,
          "enumeration": null,
          "null_count": 992129,
          "content_type": "category",
          "examples": "(NULL)… [992129]\nOther [4091]\nOther… [2006]\nJFK Airport [562]\nIntersection [486]"
        },
        {
          "name": "Bridge Highway Name",
          "type": "String",
          "label": "Bridge/Highway Identifier",
          "description": "Name of the bridge or highway involved in the incident.",
          "min": "145th St. Br - Lenox Ave",
          "max": "Willis Ave Br - 125th St/1st Ave",
          "cardinality": 68,
          "enumeration": null,
          "null_count": 997711,
          "content_type": "category",
          "examples": "(NULL)… [997711]\nOther… [779]\nBelt Pkwy [276]\nBQE/Gowanus Expwy [254]\nGrand Central Pkwy [186]"
        },
        {
          "name": "Bridge Highway Direction",
          "type": "String",
          "label": "Bridge/Highway Direction",
          "description": "Travel direction for the bridge or highway (e.g., Eastbound, Westbound).",
          "min": "Bronx Bound",
          "max": "Westbound/To Goethals Br",
          "cardinality": 50,
          "enumeration": null,
          "null_count": 997691,
          "content_type": "category",
          "examples": "(NULL)… [997691]\nOther… [987]\nEast/Long Island Bound [210]\nNorth/Bronx Bound [208]\nEast/Queens Bound [197]"
        },
        {
          "name": "Road Ramp",
          "type": "String",
          "label": "Ramp/Road Indicator",
          "description": "Indicates whether the incident involved a road ramp, roadway, or is not applicable.",
          "min": "N/A",
          "max": "Roadway",
          "cardinality": 4,
          "enumeration": null,
          "null_count": 997693,
          "content_type": "category",
          "examples": "(NULL)… [997693]\nRoadway [1731]\nRamp [555]\nN/A [21]"
        },
        {
          "name": "Bridge Highway Segment",
          "type": "String",
          "label": "Bridge/Highway Segment ID",
          "description": "Specific segment identifier of the bridge or highway where the incident occurred.",
          "min": "1-1-1265963747",
          "max": "Wythe Ave/Kent Ave (Exit 31)",
          "cardinality": 937,
          "enumeration": null,
          "null_count": 997556,
          "content_type": "category",
          "examples": "(NULL)… [997556]\nOther… [2144]\nRamp [92]\nRoadway [54]\nClove Rd/Richmond Rd (Exi… [23]"
        },
        {
          "name": "Latitude",
          "type": "Float",
          "label": "Geographic Latitude",
          "description": "Decimal‑degree latitude of the incident location, used for spatial analysis.",
          "min": "40.1123853",
          "max": "40.9128688",
          "cardinality": 353694,
          "enumeration": null,
          "null_count": 254695,
          "content_type": "latitude",
          "examples": "Other… [739329]\n(NULL)… [254695]\n40.89187241649303 [1538]\n40.1123853 [1153]\n40.89238451539139 [663]"
        },
        {
          "name": "Longitude",
          "type": "Float",
          "label": "Geographic Longitude",
          "description": "Decimal‑degree longitude of the incident location.",
          "min": "-77.5195844",
          "max": "-73.7005968",
          "cardinality": 353996,
          "enumeration": null,
          "null_count": 254695,
          "content_type": "longitude",
          "examples": "Other… [739329]\n(NULL)… [254695]\n-73.86016845296459 [1538]\n-77.5195844 [1153]\n-73.8592161325675 [663]"
        },
        {
          "name": "Location",
          "type": "String",
          "label": "Coordinate Tuple",
          "description": "String representation of latitude and longitude as a tuple, e.g., \"(40.8919, -73.8602)\".",
          "min": "(40.1123853, -77.5195844)",
          "max": "(40.9128688, -73.9024731)",
          "cardinality": 375772,
          "enumeration": null,
          "null_count": 254695,
          "content_type": "free_text",
          "examples": "Other… [739329]\n(NULL)… [254695]\n(40.89187241649303, -73.8… [1538]\n(40.1123853, -77.5195844) [1153]\n(40.89238451539139, -73.8… [663]"
        }
      ],
      "enum_threshold": 10,
      "num_examples": 5,
      "truncate_str": 25,
      "attribution": "Generated by qsv v20.1.0 describegpt\nCommand line: target/debug/qsv describegpt --all --two-pass /tmp/NYC_311_SR_2010-2020-sample-1M.csv -o /tmp/nyc311-describegpt.json --fresh --infer-content-type --format json\nPrompt file: Default v7.1.0\nModel: openai/gpt-oss-20b\nLLM API URL: http://localhost:1234/v1\nLanguage: \nTimestamp: 2026-05-30T00:55:23.334737+00:00\n\nWARNING: Label and Description generated by an LLM and may contain inaccuracies. Verify before using!\n",
      "relationships": [
        {
          "kind": "ordered",
          "members": [
            "Created Date",
            "Closed Date"
          ],
          "anchor": "Created Date"
        },
        {
          "kind": "correlated",
          "members": [
            "Latitude",
            "Longitude"
          ]
        }
      ]
    },
    "reasoning": "FIRST PASS REASONING:\nWe need to generate labels, descriptions, and content_type for all 41 fields. Use summary stats and frequency distribution.\n\nWe must choose a content type from allowed tokens: first_name, last_name, full_name, username, password, email, phone, street_address, street_name, building_number, secondary_address, city, state, state_abbr, zip_code, country, country_code, latitude, longitude, time_zone, company_name, industry, job_title, profession, unique_id, uuid, credit_card, currency_code, isbn, ip_address, ipv6_address, mac_address, url, user_agent, file_name, file_path, mime_type, color_hex, license_plate, date, datetime, time, duration, category, lorem_word, lorem_sentence, lorem_paragraph, free_text, unknown. plus suffix for date/datetime: :<fmt>. duration may have :N.\n\nWe must infer content type based on field semantics and data. For many fields we can assign 'city', 'state_abbr' etc.\n\nLet's list each field:\n\n1. Unique Key - integer unique ID; cardinality 1,000,000 all unique -> unique_id? but rule: if unique (all unique), deterministic override will set unique_id regardless of our content_type. We cannot use unique_id ourselves because it's reserved; we should pick a more specific type that fits: maybe \"uuid\" or \"unknown\". Since Unique Key is an integer primary key, not a UUID string. It might be some numeric ID but not necessarily UUID. Allowed tokens: uuid only for UUID format. unknown seems safe. But could be 'unique_id' would be overridden; we should choose e.g., \"unknown\". But maybe choose \"uuid\"? The dataset likely uses integer incremental IDs; content type \"unknown\" is fine.\n\n2. Created Date - datetime with format \"01/24/2013 12:00:00 AM\" etc. So content_type: datetime:%m/%d/%Y %I:%M:%S %p\n\n3. Closed Date - similar format, but range includes year 2100 etc. same format.\n\n4. Agency - string like \"NYPD\". This is an agency abbreviation or code; could be category? maybe company_name? But these are agencies. Use \"category\" perhaps? It's a categorical field. Allowed token: category. So content_type: category\n\n5. Agency Name - full names, e.g., \"New York City Police Department\". Could be \"company_name\" or \"full_text\"? Use company_name.\n\n6. Complaint Type - categories like Noise - Residential. Use category.\n\n7. Descriptor - descriptor of complaint; also categorical text. Use category? Might be free_text? But descriptors are specific phrases, but still categorical: e.g., Loud Music/Party. Probably category.\n\n8. Location Type - type like RESIDENTIAL BUILDING. Category.\n\n9. Incident Zip - zip codes string length 5. So zip_code\n\n10. Incident Address - street address string. Use street_address? But address may include cross streets; but typical format \"655 EAST 230 STREET\". Street address: use street_address.\n\n11. Street Name - street name. Use street_name\n\n12. Cross Street 1 - street names. street_name\n\n13. Cross Street 2 - same.\n\n14. Intersection Street 1 - street_name\n\n15. Intersection Street 2 - street_name\n\n16. Address Type - values like ADDRESS, INTERSECTION, BLOCKFACE, LATLONG, PLACENAME. Category: category\n\n17. City - city names; use city\n\n18. Landmark - string maybe street name or landmark; could be street_name? But landmarks are also addresses. Use free_text? But better to treat as free_text or category? Landmarks like \"EAST 230 STREET\" etc. Could be street_name? Actually these are street addresses or names, but might not be official street names. Maybe use free_text.\n\n19. Facility Type - values like DSNY Garage, School District. Use category\n\n20. Status - status of complaint; category\n\n21. Due Date - datetime format similar to Created/Closed date: 04/08/2015 10:00:58 AM. So datetime:%m/%d/%Y %I:%M:%S %p\n\n22. Resolution Description - free text sentences, long. Use free_text\n\n23. Resolution Action Updated Date - datetime same format.\n\n24. Community Board - values like \"0 Unspecified\" etc. Category? But it's a numeric board ID + name. Could be category or free_text. Use category.\n\n25. BBL - borough block lot number: 10-digit string maybe numeric. Use unknown? It's a unique identifier for NYC property. Use unknown.\n\n26. Borough - borough names like BRONX, QUEENS. state_abbr? Actually boroughs are not states; but we can use \"state\" or \"city\"? Allowed tokens: city, state, state_abbr. But borough is similar to city. We could use city. So content_type: city.\n\n27. X Coordinate (State Plane) - integer coordinate. Could be unknown numeric field. Use unknown.\n\n28. Y Coordinate (State Plane) - same; unknown.\n\n29. Open Data Channel Type - categories like PHONE, UNKNOWN, ONLINE, MOBILE, OTHER. category\n\n30. Park Facility Name - names of park facilities; category? or free_text. It's a name string; maybe category.\n\n31. Park Borough - borough names again; city type.\n\n32. Vehicle Type - categories: Car Service, Ambulette / Paratransit etc. category\n\n33. Taxi Company Borough - borough names again; city.\n\n34. Taxi Pick Up Location - strings like Other, JFK Airport, Intersection. Category.\n\n35. Bridge Highway Name - highway names. category.\n\n36. Bridge Highway Direction - direction categories. category.\n\n37. Road Ramp - values: Roadway, Ramp, N/A. category.\n\n38. Bridge Highway Segment - segment descriptions; category.\n\n39. Latitude - numeric float coordinates. Use latitude\n\n40. Longitude - numeric float coordinates. Use longitude\n\n41. Location - coordinate pair string like \"(40.89187241649303, -73.86016845296459)\". Might be free_text or location coordinate representation. Could use free_text but better maybe \"free_text\" because it's a tuple.\n\nNow produce JSON with fields mapping to label, description, content_type.\n\nNeed to include relationships array: identify inter-column relationships. Likely:\n\n- created_date <= closed_date (ordered). So kind ordered, members [\"Created Date\",\"Closed Date\"], anchor \"Created Date\".\n\n- maybe due date relative to created? Not necessarily monotonic but could be after created? Might not guarantee.\n\n- Latitude and Longitude are correlated; but we can treat as correlated group. But only one relationship per field. We can include a correlated relationship: members [\"Latitude\",\"Longitude\"]. Anchor none.\n\n- Maybe BBL, Borough, City, Incident Zip, Location Type etc might have joint relationships but we can't be sure. Could define joint for borough and city? Actually city includes borough names as part of larger city names; but maybe not necessary.\n\n- Also X Coordinate and Y Coordinate correlated with Latitude/Longitude? But we cannot create multiple groups overlapping fields; rule: \"Do NOT place a field in more than one relationship.\" So if we include latitude & longitude correlated, we can't include them elsewhere. That's fine.\n\nWe might also have joint relationship for Incident Zip, City, Borough? They are related but we can choose one group: maybe [\"Incident Zip\",\"City\"] as joint? But they belong to same location; but may be independent? For simplicity, only include relationships we are confident: ordered (created_date <= closed_date) and correlated (Latitude,Longitude). We might also consider correlated between X Coordinate (State Plane) and Y Coordinate. They correlate with each other; but then we would have field used in two relationships if we also use Latitude/Longitude. But rule says no overlapping. So choose only one relationship for coordinate pairs: either latitude-longitude or x/y coordinates. Which is more relevant? Probably lat/lon is standard, so correlated group of latitude & longitude.\n\nThus relationships array with those two groups.\n\nNow produce JSON.\n\nWe need to provide label same as field names but maybe human-friendly (e.g., \"Unique Key\" etc). Description: full description.\n\nLet's craft descriptions for each. Use information from summary and freq distribution.\n\nWe'll create concise but multiple sentences.\n\nAlso ensure content_type tokens exactly match allowed list. Add suffix for datetime format.\n\nNeed to confirm the datetime format string. Frequency shows date values like \"01/24/2013 12:00:00 AM\". Format: mm/dd/yyyy hh:mm:ss AM/PM. So strftime: \"%m/%d/%Y %I:%M:%S %p\"\n\nAlso check that due dates also same format.\n\nNow produce JSON object with keys as field names exactly as given in list. Use quotes.\n\nLet's craft each description:\n\n1. Unique Key: \"A unique integer identifier for each record, serving as a surrogate primary key.\"\n\n2. Created Date: \"The timestamp when the complaint was first recorded in the system.\"\n\n3. Closed Date: \"The timestamp indicating when the complaint case was closed or resolved.\"\n\n4. Agency: \"Abbreviation of the agency responsible for handling the complaint (e.g., NYPD, DOT).\"\n\n5. Agency Name: \"Full name of the agency that processed the complaint.\"\n\n6. Complaint Type: \"Broad category describing the nature of the complaint, such as noise or parking issues.\"\n\n7. Descriptor: \"Specific description providing more detail about the complaint type.\"\n\n8. Location Type: \"Classification of the physical location where the incident occurred (e.g., residential building, street).\"\n\n9. Incident Zip: \"Five-digit ZIP code identifying the area of the incident.\"\n\n10. Incident Address: \"Street address or intersection where the incident took place.\"\n\n11. Street Name: \"Primary street name involved in the incident.\"\n\n12. Cross Street 1: \"First cross street intersecting at the incident location.\"\n\n13. Cross Street 2: \"Second cross street intersecting at the incident location, if applicable.\"\n\n14. Intersection Street 1: \"One of the streets forming an intersection at the incident site.\"\n\n15. Intersection Street 2: \"Other street forming the intersection at the incident site.\"\n\n16. Address Type: \"Code indicating the type of address entry (e.g., ADDRESS, INTERSECTION).\"\n\n17. City: \"City or borough name where the incident occurred.\"\n\n18. Landmark: \"Notable landmark or feature near the incident location.\"\n\n19. Facility Type: \"Type of facility involved, such as a garage or school district.\"\n\n20. Status: \"Current status of the complaint (e.g., Closed, Pending).\"\n\n21. Due Date: \"Deadline date and time by which the agency is expected to resolve the complaint.\"\n\n22. Resolution Description: \"Narrative detailing actions taken or outcomes for the complaint.\"\n\n23. Resolution Action Updated Date: \"Timestamp of the most recent update to the resolution action record.\"\n\n24. Community Board: \"Identifier and name of the community board overseeing the area.\"\n\n25. BBL: \"Borough‑Block‑Lot number uniquely identifying a NYC property parcel.\"\n\n26. Borough: \"Name of the borough where the incident took place.\"\n\n27. X Coordinate (State Plane): \"X coordinate in the State Plane coordinate system for the incident location.\"\n\n28. Y Coordinate (State Plane): \"Y coordinate in the State Plane coordinate system for the incident location.\"\n\n29. Open Data Channel Type: \"Channel through which the complaint was submitted, e.g., PHONE or ONLINE.\"\n\n30. Park Facility Name: \"Name of a park facility related to the incident, if applicable.\"\n\n31. Park Borough: \"Borough where the referenced park facility is located.\"\n\n32. Vehicle Type: \"Type of vehicle involved in the incident, such as Car Service or Taxi.\"\n\n33. Taxi Company Borough: \"Borough where the taxi company operates.\"\n\n34. Taxi Pick Up Location: \"Location from which a taxi was picked up for the incident.\"\n\n35. Bridge Highway Name: \"Name of the bridge or highway at the incident site.\"\n\n36. Bridge Highway Direction: \"Direction designation for the bridge or highway (e.g., Eastbound).\"\n\n37. Road Ramp: \"Indicates whether the incident involved a road ramp, roadway, or N/A.\"\n\n38. Bridge Highway Segment: \"Specific segment identifier of the bridge or highway.\"\n\n39. Latitude: \"Geographic latitude coordinate of the incident location in decimal degrees.\"\n\n40. Longitude: \"Geographic longitude coordinate of the incident location in decimal degrees.\"\n\n41. Location: \"Combined latitude and longitude coordinates formatted as a string tuple.\"\n\nNow content types.\n\nLet's confirm each mapping:\n\n1. Unique Key -> unknown\n\n2. Created Date -> datetime:%m/%d/%Y %I:%M:%S %p\n\n3. Closed Date -> same\n\n4. Agency -> category\n\n5. Agency Name -> company_name\n\n6. Complaint Type -> category\n\n7. Descriptor -> category\n\n8. Location Type -> category\n\n9. Incident Zip -> zip_code\n\n10. Incident Address -> street_address\n\n11. Street Name -> street_name\n\n12. Cross Street 1 -> street_name\n\n13. Cross Street 2 -> street_name\n\n14. Intersection Street 1 -> street_name\n\n15. Intersection Street 2 -> street_name\n\n16. Address Type -> category\n\n17. City -> city\n\n18. Landmark -> free_text\n\n19. Facility Type -> category\n\n20. Status -> category\n\n21. Due Date -> datetime:%m/%d/%Y %I:%M:%S %p\n\n22. Resolution Description -> free_text\n\n23. Resolution Action Updated Date -> datetime:%m/%d/%Y %I:%M:%S %p\n\n24. Community Board -> category\n\n25. BBL -> unknown\n\n26. Borough -> city (or maybe state? but borough is a subregion; we can use city)\n\n27. X Coordinate (State Plane) -> unknown\n\n28. Y Coordinate (State Plane) -> unknown\n\n29. Open Data Channel Type -> category\n\n30. Park Facility Name -> category\n\n31. Park Borough -> city\n\n32. Vehicle Type -> category\n\n33. Taxi Company Borough -> city\n\n34. Taxi Pick Up Location -> category\n\n35. Bridge Highway Name -> category\n\n36. Bridge Highway Direction -> category\n\n37. Road Ramp -> category\n\n38. Bridge Highway Segment -> category\n\n39. Latitude -> latitude\n\n40. Longitude -> longitude\n\n41. Location -> free_text\n\nNow relationships:\n\nordered: [\"Created Date\",\"Closed Date\"], anchor \"Created Date\"\n\ncorrelated: [\"Latitude\",\"Longitude\"]\n\nAlso maybe correlated for X Coordinate and Y coordinate? But we cannot overlap; choose lat/long.\n\nReturn JSON accordingly.\n\nLet's craft final output.\n\nREFINE PASS REASONING:\nWe need to produce refined dictionary for all 41 fields. We must consider composite concepts like mailing address, coordinates, dates etc.\n\nAlso content_type: For Unique Key we omit content_type because it's unique_id.\n\nFor others: need to choose best token.\n\nLet's list fields and propose new label, description, content_type.\n\nWe'll keep first-pass values but refine.\n\nWe need cross-field relationships:\n\nMailing Address components: Incident Zip, Incident Address, Street Name, Cross Streets, Intersection streets, City, Landmark? Also Address Type indicates type. So we can say Street Name is component of incident address; maybe rename to \"Incident Street Name\". But label may remain same but mention it's part.\n\nCoordinates: Latitude, Longitude, X/Y coordinate state plane, Location string; combine into geographic coordinates.\n\nDates: Created Date, Closed Date, Due Date, Resolution Action Updated Date. Also date ranges? We might not combine.\n\nAgency fields: Agency abbreviation and Agency Name - keep.\n\nComplaint Type, Descriptor: maybe group as Complaint Category.\n\nLocation Type, Facility Type, Address Type etc.\n\nCommunity Board, Borough, BBL.\n\nVehicle type, Taxi company borough, taxi pick up location.\n\nBridge Highway fields.\n\nPark facility name, Park borough.\n\nOk.\n\nNow produce JSON with field names as keys. Must follow same shape: \"fields\": { ... }? The example output in instructions says:\n\n{\n  \"field_name_1\": {\n    \"label\": \"...\",\n    ...\n  },\n  ...\n}\n\nSo root object mapping field names to objects. Not array.\n\nWe need to provide for each of the 41 fields a dict with label, description, content_type if not omitted.\n\nAlso we should keep same keys: \"label\", \"description\", \"content_type\". The first-pass had other attributes but we don't include them.\n\nOk.\n\nLet's craft entries.\n\nUnique Key: omit content_type. Provide label \"Record ID\" maybe? Keep as Unique Key.\n\nDescription: unique integer identifier for each complaint record, surrogate primary key.\n\nContent type omitted.\n\nCreated Date: date:%m/%d/%Y . But first-pass had timezone? It shows \"01/01/2010\". So it's date only. Use content_type \"date:%m/%d/%Y\".\n\nClosed Date: also date:%m/%d/%Y.\n\nDue Date: datetime:%m/%d/%Y %I:%M:%S %p.\n\nResolution Action Updated Date: date:%m/%d/%Y? Actually first-pass shows \"12/31/2009\" to \"12/23/2020\". So date only. Use date:%m/%d/%Y.\n\nAgency: category token maybe. Content type \"category\".\n\nAgency Name: company_name.\n\nComplaint Type: category.\n\nDescriptor: free_text or category? It's more detailed descriptor; could be category, but many unique values. We'll keep category? But first-pass had content_type category. Keep category.\n\nLocation Type: category.\n\nIncident Zip: zip_code.\n\nIncident Address: street_address.\n\nStreet Name: street_name (component). Provide description referencing incident address.\n\nCross Street 1 and Cross Street 2: street_name components.\n\nIntersection Street 1/2: street_name components.\n\nAddress Type: category.\n\nCity: city.\n\nLandmark: free_text.\n\nFacility Type: category.\n\nStatus: category.\n\nDue Date etc already.\n\nResolution Description: free_text.\n\nCommunity Board: category. Might keep category.\n\nBBL: unknown? In first-pass content_type \"unknown\". Keep unknown or maybe treat as string but unknown.\n\nBorough: city (but borough names). Use city token.\n\nX Coordinate (State Plane): unknown. Keep unknown.\n\nY Coordinate (State Plane): unknown.\n\nOpen Data Channel Type: category.\n\nPark Facility Name: category.\n\nPark Borough: city.\n\nVehicle Type: category.\n\nTaxi Company Borough: city.\n\nTaxi Pick Up Location: category.\n\nBridge Highway Name: category.\n\nBridge Highway Direction: category.\n\nRoad Ramp: category.\n\nBridge Highway Segment: category.\n\nLatitude: latitude.\n\nLongitude: longitude.\n\nLocation: free_text (string of tuple). Maybe could be \"unknown\" but keep free_text.\n\nOk.\n\nWe need to refine descriptions referencing cross-field relationships where relevant.\n\nLet's craft for each field.\n\nWill produce final JSON.",
    "token_usage": {
      "prompt": 22906,
      "completion": 7760,
      "total": 30666,
      "elapsed": 112892
    }
  },
  "Description": {
    "response": "**Description**\n\nThe dataset is a one‑million‑record sample of New York City 311 service requests collected between January 2010 and December 2020. Each record represents an individual complaint, uniquely identified by the surrogate key *Unique Key*. The primary temporal fields—*Created Date* and *Closed Date*—span roughly a decade, with most complaints created in 2013‑2015 and closed shortly thereafter. Complaint content is encoded through several hierarchical categories (*Agency*, *Complaint Type*, *Descriptor*) and geospatial identifiers (*Latitude/Longitude*, *Borough*, *BBL*). The data also includes administrative metadata such as the submitting channel, resolution narrative, and community board jurisdiction.\n\n**Notable Characteristics**\n\n- **Temporal concentration**: Created dates cluster around 2013‑2015 (≈34 % of records) with a mild negative skew (-0.017), indicating a slight leftward tail in complaint volume.\n- **High cardinality & uniqueness**: Latitude/Longitude and BBL have >35,000 unique values each; <ALL_UNIQUE> entries dominate the frequency tables, implying fine‑grained spatial granularity but sparse categorical support for modeling.\n- **Dominant “Other” categories**: For *Complaint Type* (~56 % of cases) and *Descriptor* (~67 %) the value “Other” absorbs a large share of complaints, signalling poor classification coverage and potential data quality issues.\n- **Missingness patterns**:  \n  - *Closed Date*: 28,619 nulls (≈2.9 %).  \n  - *Incident Zip*, *Address Type*, *Due Date*: 55k–126k nulls each (~5‑12 %).  \n  - *Resolution Description* has 20,480 nulls (≈2 %).\n- **Duplicate handling**: The surrogate primary key ensures no duplicate records; however, many fields exhibit duplicate values across records (e.g., Agency codes, borough names), which is expected for categorical data.\n- **Outliers & placeholders**: *Closed Date* min/max values (1900‑01‑01 / 2100‑01‑01) appear to be placeholder defaults rather than real dates; these should be treated as missing or censored.  \n  Similarly, *Latitude/Longitude* extremes (e.g., –77.5196, –73.7006) correspond to boundary coordinates of the NYC metro area.\n- **PII / PHI**: The dataset contains street addresses, ZIP codes, and latitude/longitude pairs that can be used to re‑identify individuals or properties; it should not be shared without appropriate de‑identification or aggregation.  \n  No PCI information is present.\n- **Data sparsity & sparsity ratio**: Fields such as *Resolution Action Updated Date* (≈690k distinct values) and *Vehicle Type* have many nulls, leading to sparse columns that may require dimensionality reduction for modeling.\n- **Relational structure**: The provided relationships indicate an ordered dependency between *Created Date* and *Closed Date*, and a spatial correlation between *Latitude* and *Longitude*.\n\n---\n\nGenerated by qsv v20.1.0 describegpt\nCommand line: target/debug/qsv describegpt --all --two-pass /tmp/NYC_311_SR_2010-2020-sample-1M.csv -o /tmp/nyc311-describegpt.json --fresh --infer-content-type --format json\nPrompt file: Default v7.1.0\nModel: openai/gpt-oss-20b\nLLM API URL: http://localhost:1234/v1\nLanguage: \nTimestamp: 2026-05-30T00:56:03.982714+00:00\n\nWARNING: Description generated by an LLM and may contain inaccuracies. Verify before using!\n",
    "reasoning": "We need to generate a description for the dataset as a whole, using summary statistics and frequency distribution. Not to output per-field data. Provide 1-8 sentences description. Then \"Notable Characteristics\" bulleted list with key characteristics like central tendency, spread, anomalies, patterns; mention outliers, missing values, duplicates, PII/PHI/PCI; other data quality issues.\n\nWe have summary statistics and frequency distribution: fields include Unique Key (unique id), Created Date (date range 2010-2020, many distinct dates with some concentration), Closed Date (range 1900-2100 but many nulls). Agency etc. We should mention dataset is NYC 311 complaint data from 2010-2020 sample of 1M records.\n\nWe can note distribution: Created Dates heavily concentrated around 2013-2015, skewness negative small -0.0169; Closed Date similar.\n\nComplaint Type has heavy skew with \"Other\" dominating (~56%). Descriptor also dominated by Other (67%).\n\nLocation types: Residential building most common.\n\nIncident Zip: many missing? Actually Incident Zip has nulls 54978 out of 1M (5%?). Actually cardinality 535, nullcount 54978 ~5.5%.\n\nAddress Type: large null counts.\n\nLatitude/Longitude: many unique values, high cardinality but distribution fairly concentrated in NYC area; there is a \"Other\" category for coordinates with >99% others? Wait Frequency shows <ALL_UNIQUE> for Latitude and Longitude (not aggregated). The frequency distribution indicates that the majority of coordinates are unique (<ALL_UNIQUE>) 100%? Actually Frequency Distribution for Latitude has value \"<ALL_UNIQUE>\" count=353694 out of 1M, meaning many unique values. So we can mention high uniqueness.\n\nWe also have duplicates: Unique Key is surrogate primary key; all unique.\n\nMissing data: Many fields with null counts: Closed Date (28k), Incident Zip (~55k), Address Type (~126k), etc.\n\nOutliers: Created Date min 2010-01-01, max 2020-12-23. Closed Date range 1900 to 2100 but many default values? The min and max represent placeholders maybe; actual closed dates probably within realistic range.\n\nWe should mention PII/PHI/PCI: Data includes addresses, coordinates, agency names. Contains personally identifiable info like addresses. It's not PCI.\n\nData quality issues: Many \"Other\" categories indicating lack of standardization; high cardinality for certain fields leading to sparse data; missing values.\n\nAlso note that the dataset includes \"Due Date\" with many nulls (647k). So we can mention.\n\nAlso note that there are multiple date columns, some may have overlapping or duplicates.\n\nWe should also mention relationships: Created Date precedes Closed Date.\n\nNow produce description 1-8 sentences. Then bullet list.\n\nAdd attribution placeholder {GENERATED_BY_SIGNATURE}.\n\nLet's craft.\n\n",
    "token_usage": {
      "prompt": 20089,
      "completion": 1264,
      "total": 21353,
      "elapsed": 40646
    }
  },
  "Tags": {
    "response": {
      "tags": [
        "nyc_311_complaints",
        "public_service_requests",
        "geocoded_incident_location",
        "complaint_resolution_status",
        "city_government_agencies",
        "noise_complaint",
        "parking_violation",
        "boroughs_of_nyc",
        "temporal_data",
        "multiagency_response"
      ],
      "attribution": "Generated by qsv v20.1.0 describegpt\nCommand line: target/debug/qsv describegpt --all --two-pass /tmp/NYC_311_SR_2010-2020-sample-1M.csv -o /tmp/nyc311-describegpt.json --fresh --infer-content-type --format json\nPrompt file: Default v7.1.0\nModel: openai/gpt-oss-20b\nLLM API URL: http://localhost:1234/v1\nLanguage: \nTimestamp: 2026-05-30T00:56:55.421583+00:00\n\nWARNING: Tags generated by an LLM and may contain inaccuracies. Verify before using!\n"
    },
    "reasoning": "We need to produce Tags (max 10) in JSON format. Use lowercase and underscore separated words, no field names. Need to consider dataset contents: NYC 311 complaints, with categories like noise, parking, etc. We should derive tags from summary stats and freq distribution. The dataset covers 2010-2020 NYC 311 complaints.\n\nWe need tags that describe the content: e.g., \"nyc_311_complaints\", \"noise_complaint\", \"parking_issue\", \"public_safety\", \"city_services\", \"complaint_resolution\", \"geospatial_data\", \"boroughs\", \"time_series\", \"incident_location\". We can pick 10.\n\nLet's think of tags in descending importance. The dataset is about NYC 311 complaints, covering many agencies and categories. So top tag: nyc_311_complaints. Next: noise_and_sounds? Actually noise is major category but not only. But we want general tags. Maybe \"public_service_requests\". Another: \"complaint_data\". \"city_government\". \"incident_reporting\". \"geocoded_location\" (since lat/lon). \"time_series\". \"borough_data\".\n\nBut we need to think of 10 unique tags, lowercase underscore separated.\n\nPossible tags list:\n\n- nyc_311_complaints\n- public_service_requests\n- noise_and_sounds? maybe \"noise_complaint\".\n- parking_violation\n- city_government_agencies\n- geocoded_incident_location\n- boroughs_of_nyc\n- time_series_data\n- complaint_resolution_status\n- multiagency_response\n\nBut we should ensure they reflect the dataset content. The summary shows many agencies: NYPD, HPD, DOT, DSNY, etc. Many complaint types: noise residential, heat/hot water, illegal parking, blocked driveway, street condition, street light condition, heating, plumbing, water system, noise street sidewalk.\n\nAlso there are fields like 'Status', 'Due Date', 'Resolution Description'.\n\nGeospatial data: lat/lon, x/y coordinates, incident address etc.\n\nTime series: Created Date and Closed Date ranges 2010-2020. So \"time_series\" or \"temporal_data\".\n\nLet's craft tags in descending importance:\n\n1. nyc_311_complaints\n2. public_service_requests\n3. geocoded_incident_location\n4. complaint_resolution_status\n5. city_government_agencies\n6. noise_and_sounds (maybe combine noise_complaint)\n7. parking_violation\n8. boroughs_of_nyc\n9. temporal_data\n10. multiagency_response\n\nBut we need to ensure tags are unique and lowercase underscore separated.\n\nLet's produce JSON:\n\n{\n  \"tags\": [\n    \"nyc_311_complaints\",\n    \"public_service_requests\",\n    \"geocoded_incident_location\",\n    \"complaint_resolution_status\",\n    \"city_government_agencies\",\n    \"noise_complaint\",\n    \"parking_violation\",\n    \"boroughs_of_nyc\",\n    \"temporal_data\",\n    \"multiagency_response\"\n  ],\n  \"attribution\": \"{GENERATED_BY_SIGNATURE}\"\n}\n\nCheck that we didn't use field names. We used 'complaint_resolution_status' not referencing fields.\n\nMake sure JSON is pretty-printed, with proper escaping (none needed). Provide tags array and attribution key at top level after tags? Problem statement: \"Add an Attribution with the placeholder \"{GENERATED_BY_SIGNATURE}\" after the Tags. If generating JSON format, add the Attribution as a separate key at the top level of the JSON object, after the Tags\". So tags first then attribution.\n\nThus final output should be:\n\n{\n  \"tags\": [...],\n  \"attribution\": \"{GENERATED_BY_SIGNATURE}\"\n}\n\nOk.",
    "token_usage": {
      "prompt": 20076,
      "completion": 875,
      "total": 20951,
      "elapsed": 35606
    },
    "num_tags": 10,
    "tag_vocab": null
  }
}