Start

Relifold documentation

Build, review, improve, and operate semantic data pipelines.

10-minute path Build your first pipeline

Connect a dataset, create a version, run it, and inspect the session.

Reference Choose an operation

Find the right deterministic or semantic operation for the result you need.

Quality loop Review and improve results

Create checkpoints, curate examples, and carry accepted improvements forward.

Learn by workflow

Bring data into SpacesUploads, S3, cloud storage, and PostgreSQL Plan from a desired outcomeGenerate an editable pipeline draft from an instruction Research one local decision systemGroup a semantic task with the steps that prepare it, then measure quality and cost Learn Relifold's hybrid methodUse hard pair gates, semantic judgment, and deterministic predecessors Research better candidate rulesCompare lossy blockers, marginal sets, compositions, and registered parameters Run and observeProgress, cost, calls, output, and incremental refresh Work through PythonUse the same public product surface as the browser

How the documentation is organized

Tutorials walk through a complete outcome. Concept pages explain durable product ideas. Operator pages document parameters and worked examples. How-to pages cover reviews, deployment, teams, and troubleshooting.