Skip to main content
WiseRegent

Lineage & Provenance

Trace every model back to its source.

Know exactly what data trained your model, what features shaped its behavior, and every service that depends on it in production — all tracked automatically.

Automated extraction

Lineage discovered from your ML frameworks, pipelines, and APIs — no manual tagging required.

Full provenance graph

See every upstream source and every downstream consumer in a single interactive view.

Impact analysis

Understand the blast radius of any data change, model update, or policy violation before it happens.

Know exactly what went into your models.

WiseRegent automatically traces the path from raw data to trained model — capturing every transformation, feature engineering step, training run, and evaluation. When a data issue surfaces, you know immediately which models are affected.

  • Training data lineage: raw sources → preprocessing → feature store → training run.
  • Experiment tracking: links MLflow/W&B runs to the resulting model artifact.
  • Automatic propagation: lineage updates as you retrain, fine-tune, or update data pipelines.

Data Lineage — fraud-detection-v3

Complete
Raw sourcetransactions.parquet
Preprocessingfraud-pipeline v2.1
Feature storefeature-store-v2
Training labelslabels-v1
Training runmlflow://run/8a3f...
Evaluation setholdout-2026-q1

See what breaks before you make a change.

Before retiring a dataset, updating a model, or changing a feature pipeline, you need to know what's downstream. WiseRegent's impact analysis shows every dependent model, API, product, and service — so changes are deliberate, not accidental.

  • Downstream mapping: every API, microservice, and product feature that depends on your model.
  • Change impact scoring: assess risk before deploying updates or retiring assets.
  • Dependency graph export: share impact reports with platform teams and stakeholders.

Downstream Impact — fraud-data-v2

3 downstream
fraud-detection-v3ML Model — HIGH risk
customer-risk-scoreML Model — MED risk
transaction-api-v2Endpoint — production
Impact scoreCritical

Lineage captured automatically — no annotation required.

01

Connect

Authorize your ML platforms, data warehouses, and pipelines.

02

Extract

WiseRegent extracts lineage from framework metadata and run logs.

03

Visualize

Full provenance graph — upstream and downstream — rendered live.

04

Analyze

Run impact analysis before every change or retirement.

Auto-extracted from your existing tools

MLflowWeights & BiasesKubeflow PipelinesSageMaker ExperimentsVertex AI MetadataUnity CatalogApache AirflowdbtFeastDelta LakeOpenLineage

Early access open

Govern your AI agents before your auditors do.

Join the early-access programme. Shape the roadmap. Lock in founding rates. No credit card, no commitment.

SOC 2 Type IIGDPR-readyNo credit cardRespond in 1–2 days