Skip to main content

H3 Data Retrieval Service

Using an AI agent?

Copy this file and pass it to your AI agent to partition, index and query your data. Or paste this prompt to get a canvas:

Make sure the Fused CLI is installed. Read https://docs.fused.io/skills/H3.md and https://github.com/fusedio/skills. Then create a Workbench canvas named NYC311_H3 that partitions NYC 311 requests for 2024-07-04 into H3 on fd://, indexes them and maps requests per res-9 hex. Push it and give me its URL.

The Fused H3 Data Retrieval Service returns any area or time window of your data within 2 seconds, even for datasets with billions of points. New to H3? Start with Working with H3.

It takes three steps: partition, index, then query.

The map below queries 1.12 billion AIS ship positions (July to October 2024), ingested this way, from the GET /h3/query endpoint. Pan or change the time window, and new results load in under 2 seconds.

Open this example in Canvas → · See more in H3 Large Datasets.

To partition and index your own data, see Partition, Index and Query for the steps and an example.

Workflow​

StepWhat it doesReference
PartitionWrites your rows as H3 Parquet files, in folders by the values you filter on. Takes a source, such as a UDF that returns your rows.fused.h3.partition()
IndexRecords which cells, partition values and times each file coversfused.h3.index()
Query in a UDFReads back only the rows you ask for, by area, point, cell, time window or partition value. Runs only inside a UDF on Fused.fused.h3.query()
Query over HTTPThe same query as an endpoint that returns Parquet, for any language or web pageGET /h3/query
VisualizeDraws the dataset on a map with the Fused Map widget and a layer of type fused-hex, with no UDFH3 Visualization

Examples​

PageWhat it shows
Partition, Index and QueryBest practices for partition, index and query
H3 large datasets39.6 million NYC 311 requests, 1.12 billion AIS positions and 367.8 million LiDAR points, queried live

Reference​