Open source
- Runs on
- A Spark cluster, or one machine
- Scale
- No cap
- Terms
- Open source, BSD-3
The Products
Comparing every possible pair across large tables is impractical. Blocking quickly filters out unlikely pairs while aiming to retain true matches. Matching then uses rules or machine learning to decide whether each remaining pair refers to the same entity. We provide products for both the blocking and matching steps.
Built for speed, scale, and accuracy.
Relixa’s blocking and matching engines are designed to handle datasets ranging from millions to billions of records. They combine advanced algorithms with efficient execution to make large-scale entity matching practical and affordable.
Blocking
Sparkly
Fast, easy-to-use blocking. Choose Sparkly to start blocking quickly without designing rules yourself. It is also a good choice when blocking speed is your main concern.
| Edition | Open source | Lite | Plus | Pro |
|---|---|---|---|---|
| Runs on | A Spark cluster, or one machine | A Mac with Apple silicon | A Mac with Apple silicon | On premises or in your cloud |
| Scale | No cap | Up to 3 million rows per table | Tens of millions of rows per table | Hundreds of millions to billions of rows per table |
| Terms | Open source, BSD-3 | Free, including commercial use | Licensed | Licensed |
| Availability | View repository | View repositoryFree | Request a demo | Request a demo |
Lite
- Runs on
- A Mac with Apple silicon
- Scale
- Up to 3 million rows per table
- Terms
- Free, including commercial use
View repositoryFree
Plus
- Runs on
- A Mac with Apple silicon
- Scale
- Tens of millions of rows per table
- Terms
- Licensed
Pro
- Runs on
- On premises or in your cloud
- Scale
- Hundreds of millions to billions of rows per table
- Terms
- Licensed
Delex
Flexible, rule-driven blocking. Choose Delex when you want control over blocking rules or need to combine multiple strategies to improve recall.
| Edition | Open source | Lite | Plus | Pro |
|---|---|---|---|---|
| Runs on | A Spark cluster, or one machine | A Mac with Apple silicon | A Mac with Apple silicon | On premises or in your cloud |
| Scale | No cap | Up to 3 million rows per table | Tens of millions of rows per table | Hundreds of millions to billions of rows per table |
| Terms | Open source, BSD-3 | Free, including commercial use | Licensed | Licensed |
| Availability | View repository | View repositoryFree | Request a demo | Request a demo |
Open source
- Runs on
- A Spark cluster, or one machine
- Scale
- No cap
- Terms
- Open source, BSD-3
Lite
- Runs on
- A Mac with Apple silicon
- Scale
- Up to 3 million rows per table
- Terms
- Free, including commercial use
View repositoryFree
Plus
- Runs on
- A Mac with Apple silicon
- Scale
- Tens of millions of rows per table
- Terms
- Licensed
Pro
- Runs on
- On premises or in your cloud
- Scale
- Hundreds of millions to billions of rows per table
- Terms
- Licensed
Matching
Iris
Use this matcher on the output of blocking to predict which candidate pairs match. This matcher is trained on a set of labeled examples (typically a few hundreds). It will help you quickly find examples to label if you do not yet have them.
| Edition | Open source | Plus | Pro |
|---|---|---|---|
| Runs on | One machine, or a Spark cluster | A Mac with Apple silicon | On premises or in your cloud |
| Scale | No cap | Candidate sets in the billions | Tens of billions of candidate pairs |
| Terms | Open source, BSD-3 | Licensed | Licensed |
| Availability | View repository | Request a demo | Request a demo |
Open source
- Runs on
- One machine, or a Spark cluster
- Scale
- No cap
- Terms
- Open source, BSD-3
Plus
- Runs on
- A Mac with Apple silicon
- Scale
- Candidate sets in the billions
- Terms
- Licensed
Pro
- Runs on
- On premises or in your cloud
- Scale
- Tens of billions of candidate pairs
- Terms
- Licensed