Documentation

HyperNorm Documentation

Normalize your first vendor feed in under 10 minutes.

Quickstart API Reference Connectors Core Concepts

Quickstart

Install the SDK, set your API key, define a config, and run your first pipeline. About 10 minutes from zero to normalized records.

pip install hypernorm

API Reference

Full REST API documentation for normalization, entity resolution, schema management, and authentication. Includes request and response shapes.

POST /v1/normalize

Connectors

Native connectors for Snowflake, BigQuery, Redshift, Postgres, S3, Salesforce, HubSpot, and CSV/Parquet upload.

connector: snowflake

Config Schema

Reference for the normalization config YAML structure: sources, field maps, canonical schema, entity blocking, and confidence thresholds.

canonical_schema: v1.yml

Core concepts

Understanding four concepts makes the rest of HyperNorm straightforward.

Canonical schema

The target schema your data is normalized toward. Defined as a YAML file referenced by name in each normalization config. HyperNorm validates every output record against the canonical schema before writing it downstream.

See in quickstart

Entity blocking

Before fuzzy matching, HyperNorm builds blocking keys (phonetic codes, prefix tokens) to pre-filter candidate record pairs. Without blocking, entity resolution at 50M+ records is computationally infeasible. Blocking reduces the candidate space by 99.7%.

See in API reference

Confidence threshold

Every entity match returns a confidence score (0.0 to 1.0). Matches above your configured threshold are auto-merged. Below-threshold matches are queued for human review. You control the threshold per entity type in your config.

See in API reference

Normalization rules

Declarative mappings from vendor field names to canonical field names. Includes type coercion (string to decimal), unit conversion (EUR to USD), null strategies (inherit from a primary source), and date format normalization.

See config example

Need help?

Email [email protected] or open an issue on our GitHub. We respond within one business day. We are an angel-stage team and we read every message.

Email us Open an issue