Connect CleanDims to the systems that produce and consume your dimensional data.
Source connectors observe; the reference holds; consumers read through cache libraries or the API. Bidirectional in observation, unidirectional in enforcement.
How CleanDims connects.
CleanDims reads dimensional values from source systems through connectors, holds the canonical reference centrally, and exposes the reference via API to any consumer that needs to resolve values. The connector model is bidirectional in observation but unidirectional in enforcement: CleanDims observes the values that source systems produce, but does not push the canonical back into source systems unless the source explicitly subscribes to the reference.
This is deliberate. Outbound propagation is theoretically appealing and operationally fragile: source-system APIs change without notice, picklist syncs break when vendors restructure, and the cost of maintaining outbound integrations grows without bound. Containing variance at the boundary, where the external value enters, is more honest than attempting to govern the source.
Source systems.
CleanDims connects to the systems where dimensional values originate. At launch, supported source systems include the major CRMs, ERPs, finance and HR platforms, ticketing systems, and the major data warehouses. The specific list ships with v1; the connector interface is open, so additional sources can be added by customers or by the community.
Consumer systems.
Any system that can make an HTTP request can consume the CleanDims reference. The primary consumer pattern is agents and pipelines reading through a local cache; the cache is provided as a library for the major agent and pipeline runtimes.
API.
For systems that cannot use the cache library directly, CleanDims exposes a REST API and a streaming API. The REST API is suitable for low-throughput resolution; the streaming API is suitable for keeping a custom cache implementation in sync with the reference at machine speed.
API documentation is available at docs.cleandims.com.
Custom connectors.
For source systems not on the supported list, CleanDims supports custom connectors through a documented connector interface. Custom connectors can be written in any language and run anywhere; they read values from the source, normalise the observation envelope, and post to CleanDims. The connector interface is open source; contributions from customers and the community are welcome.
A pipeline team migrating from spreadsheet mappings.
A data platform team has eight ETL pipelines that each maintain their own vendor-name mapping in a shared spreadsheet. The team installs the CleanDims Python cache library, points it at the procurement reference, and replaces the spreadsheet lookups with cleandims.resolve("vendor", raw_value) calls. The pipelines now resolve against the live reference, emit a confidence signal alongside each resolved value, and propagate cache version metadata downstream. The spreadsheet is deprecated. No new mapping diverges from anyone else's.