AgentSkillsCN

chicago-data-portal

当用户提出“查询芝加哥数据”“查找芝加哥数据集”“获取芝加哥犯罪数据”“下载芝加哥许可证”“为芝加哥编写 SODA 查询”“搜索 data.cityofchicago.org”,或提及芝加哥城市数据(311 服务、许可证、执照、检查、犯罪等)时,应优先选用此技能。

SKILL.md
--- frontmatter
name: chicago-data-portal
description: This skill should be used when the user asks to "query Chicago data", "find Chicago datasets", "get Chicago crime data", "download Chicago permits", "write a SODA query for Chicago", "search data.cityofchicago.org", or mentions Chicago city data (311, permits, licenses, inspections, crimes, etc.).
version: 1.1.0

Chicago Data Portal Skill

Query and download datasets from the City of Chicago Data Portal using the Socrata Open Data API (SODA) and SoQL.

Prerequisites

Before querying, check if the user has an app token:

  1. Look for CHICAGO_DATA_PORTAL_TOKEN in the user's .env file
  2. If found, use it in requests via header: X-App-Token: <token>
  3. If not found, instruct the user to:

Queries work without a token but are rate-limited.

Quick Start

The Chicago Data Portal is at data.cityofchicago.org. Each dataset has a unique 4x4 ID (e.g., ijzp-q8t2 for crimes). Use the catalog API to discover datasets, then query via SODA.

Workflow

Step 1: Clarify the Data Need

Ask the user:

  • Topic: What data? (crimes, permits, 311 requests, businesses, etc.)
  • Geography: Citywide, ward, community area, or specific location/radius?
  • Time window: Date range or "most recent"?
  • Output: JSON (code) or CSV (Excel)?
  • Granularity: Raw rows or aggregated counts?

Step 2: Find the Dataset

Option A - Catalog Search API:

code
GET https://api.us.socrata.com/api/catalog/v1?domains=data.cityofchicago.org&q=<keywords>

Option B - Portal UI: Browse https://data.cityofchicago.org and use the search bar.

Deliverable: Dataset name, 4x4 ID, and API endpoint.

See references/popular-datasets.md for commonly requested datasets.

Step 3: Get Dataset Metadata

Fetch schema and column info:

code
GET https://data.cityofchicago.org/api/views/<4x4-ID>

Key fields in response:

  • columns[].fieldName - exact column names for queries
  • columns[].dataTypeName - data type (text, number, calendar_date, location, etc.)
  • columns[].description - what the column means
  • rowsUpdatedAt - last data update timestamp

Always verify column names from metadata before building queries.

Step 4: Build the Query

Legacy GET (simple, recommended for most cases):

code
https://data.cityofchicago.org/resource/<4x4-ID>.json?$where=<filter>&$limit=1000

SODA3 POST (complex queries):

bash
curl -X POST \
  -H "X-App-Token: YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query": "SELECT * WHERE date > '\''2024-01-01'\''", "page": {"pageNumber": 1, "pageSize": 1000}}' \
  https://data.cityofchicago.org/api/v3/views/<4x4-ID>/query.json

Step 5: Handle Pagination

Default limit is 1000 rows. For larger extracts:

code
$limit=1000&$offset=0    # Page 1
$limit=1000&$offset=1000 # Page 2

Always include $order for stable paging:

code
$order=date DESC&$limit=1000&$offset=0

For full dataset export, use CSV:

code
https://data.cityofchicago.org/api/views/<4x4-ID>/rows.csv?accessType=DOWNLOAD

SoQL Essentials

Query Parameters

ParamPurposeExample
$selectColumns to return$select=date,primary_type,ward
$whereFilter rows$where=year=2024
$groupAggregate$group=primary_type
$havingFilter aggregates$having=count(*)>100
$orderSort results$order=date DESC
$limitMax rows$limit=500
$offsetSkip rows$offset=1000

Syntax Rules

  • Backticks around column names: `column_name`
  • Single quotes for strings: 'value'
  • Dates as ISO strings: '2024-01-01T00:00:00'

Common Filters

sql
-- Date range
$where=date >= '2024-01-01' AND date < '2025-01-01'

-- Text matching (case-insensitive)
$where=upper(primary_type) = 'THEFT'

-- Null handling
$where=ward IS NOT NULL

-- Multiple values
$where=primary_type IN ('THEFT', 'BATTERY', 'ASSAULT')

Aggregations

sql
$select=primary_type, count(*) as total
$group=primary_type
$order=total DESC

See references/soql-quick-ref.md for full function reference.

Geospatial Queries

If the dataset has a location field (Point type):

sql
-- Within radius (meters)
$where=within_circle(location, 41.8781, -87.6298, 1000)

-- Within bounding box
$where=within_box(location, 42.0, -87.9, 41.6, -87.5)

-- Within polygon
$where=within_polygon(location, 'MULTIPOLYGON(((-87.6 41.8, -87.5 41.8, -87.5 41.9, -87.6 41.9, -87.6 41.8)))')

App Tokens

Unauthenticated requests are rate-limited. Register for a free app token:

  1. Create account at https://data.cityofchicago.org
  2. Go to Developer Settings
  3. Create New App Token
  4. Use via header: X-App-Token: YOUR_TOKEN

Output Format

Provide the user with:

  1. Dataset: Name + 4x4 ID + portal link
  2. Columns used: Exact field names
  3. Query: Formatted SoQL
  4. How to run: curl command or full URL
  5. Assumptions: Time zone, update frequency, any caveats

Example Response Format

code
Dataset: Crimes - 2001 to Present (ijzp-q8t2)
https://data.cityofchicago.org/d/ijzp-q8t2

Query:
SELECT date, primary_type, description, ward, latitude, longitude
WHERE date >= '2024-01-01' AND primary_type = 'THEFT'
ORDER BY date DESC
LIMIT 100

Run it:
curl "https://data.cityofchicago.org/resource/ijzp-q8t2.json?\$select=date,primary_type,description,ward,latitude,longitude&\$where=date%20%3E=%20%272024-01-01%27%20AND%20primary_type%20=%20%27THEFT%27&\$order=date%20DESC&\$limit=100"

Note: Data updates daily. Dates are in Chicago local time (America/Chicago).

Troubleshooting

IssueFix
404 / "unknown column"Wrong dataset ID or field name. Check metadata endpoint.
Empty resultsFilters too strict, wrong date format, or nulls.
429 throttledAdd X-App-Token header.
Slow querySelect fewer columns, add filters, reduce limit.
Encoding errorsURL-encode special chars: space=%20, >=%3E, '=%27

Additional Resources

  • references/popular-datasets.md - Common Chicago datasets with IDs
  • references/soql-quick-ref.md - All SoQL functions
  • examples/python-query.py - Python code snippet
  • examples/curl-examples.sh - curl command templates