Cleora

What Cleora is

Cleora is a graph embedding engine for graph and relational data. It uses a Rust core and sparse matrix propagation to turn entities, users, items or nodes into vectors usable for recommendation, risk control, similarity retrieval and clustering.

The site emphasises CPU availability, deterministic results and no need for negative sampling or GPU clustering.

What you can do with it

• Convert nodes and relations into vectors
• Use the vectors in recommendation or retrieval models
• Analyse proximity between accounts, devices and transactions
• Reproduce embeddings across runs and versions

Who it is for

• Developers and data scientists
• Machine learning engineers
• Teams without GPU clusters

What to watch out for

• It produces features, not a system: recall, ranking, business rules and serving are separate
• You need to prepare relational data and understand what nodes and edges mean
• It is code-first, so non-engineers will need integration help
• Evaluate embeddings against your actual task, not in isolation