Understand the signed-in datasets surfaces across feature builds, contract validation, lineage, and research dataset coverage.
Use this guide when you need to move from a watchlist or research idea into a reproducible dataset artifact with contracts, lineage, and coverage context preserved.
These are the current signed-in dataset surfaces that shape the build and validation workflow.
Build feature datasets, review past runs, and manage downloadable artifacts from the main datasets page.
Select dataset contracts, validate requested columns, and confirm schema expectations before or after a build.
Inspect produced artifacts and lineage edges so the resulting dataset can be traced back through its inputs.
Fetch agent-ready context from dataset-linked instruments when the build will feed downstream automation or research.
Build research-specific universe datasets with asset-class, date, and interval controls.
Inspect coverage rows, expected bars, actual bars, and coverage percentage once a research dataset exists.
Use this sequence to keep dataset generation reproducible and reviewable.
Start by choosing the watchlist or research scope, date window, interval, and feature or universe intent.
Check the selected contract and expected columns before treating the requested artifact as valid.
Run the feature or research dataset build, then inspect the produced runs and downloadable outputs.
Review lineage edges and coverage rows so downstream consumers know what the dataset actually contains.
The dataset form is where reproducibility starts, not where it ends.
Set the watchlist, instrument, interval, and time window first so repeated builds stay comparable.
The feature-list input is part of the contract. Treat changes to it as a dataset-shape change, not just a convenience edit.
Pick the active contract before building so the requested columns and schema expectations are explicit.
The three action buttons do different jobs: build produces artifacts, validate checks schema shape, and fetch agent context packages the dataset for downstream consumers.
After a build finishes, the evidence panels determine whether the dataset is safe to reuse.
Use the contract status panel to distinguish between a reusable artifact and one that still has schema violations.
Review feature-run history so new builds are compared with prior attempts instead of treated as isolated jobs.
The dataset list is where you choose which artifact to download and which one to inspect for lineage. Keep selection intentional.
Use lineage edges and agent-context preview together when the dataset will feed research, backtests, or agent workflows.
Use these scenarios when a build needs a concrete reproducibility pattern instead of only the generic dataset loop.
Define the watchlist, date range, and feature list up front, validate against the intended contract, then compare the new run with prior feature-run history before downloading anything.
The dataset artifact is reusable because its shape, contract, and run history all agree.
Select the next contract version, run validation before a full build, and inspect contract-status or violation output before you commit a downstream scan or label workflow to the new shape.
You know whether the new schema is a safe upgrade or a breaking change before expensive rebuilds begin.
After the build succeeds, open lineage edges, choose the artifact intentionally, and fetch agent context only once the coverage rows and artifact selection match the downstream consumer.
The handoff carries provenance, coverage, and the correct artifact instead of an unverified download.
Jump into the signed-in datasets surface and review builds, contracts, artifacts, and lineage.
Use the research workflow when the dataset is part of scan, label, or workbench iteration.
Cross-check the generated contracts and endpoint surface that back dataset artifacts and workflows.
Mar 24, 2026
Report unclear guidance, stale contracts, missing coverage, or broken docs UI on this page.
Open feedback issueJump to the section you need without losing your place.
Mar 24, 2026
Report unclear guidance, stale contracts, missing coverage, or broken docs UI on this page.
Open feedback issue