File-first foundation
Bounded CSV, XLSX and supported scalar Parquet subsets are available in the trusted-local foundation; not a hosted service.
Public showcase · Registration and public uploads unavailable · AI demo not enabled.
Bring together configurable checks, source-linked findings, contextual dataset exploration, and readable reporting in a workspace designed around investigation.
Different datasets require different interpretations. The trusted-local foundation supports bounded CSV, XLSX and supported Parquet subsets; this does not qualify every format or semantic for cloud use. The website design guides users through schema interpretation and explicit analysis rules before a run. Supported explicit checks include inclusive numeric ranges, required/missingness, typed allowed values, and single/composite uniqueness. Interpretation and numeric text must remain explicit. This trusted-local foundation is not a cloud-qualified service; public analysis is unavailable.
Click a potential discrepancy to view a focused window of relevant records, a visual highlight, and an explanation. Large-dataset navigation is a proposed on-demand capability, not a claim of benchmarked production performance.
| # | order_id | age | region |
|---|
This record may be outside the configured range. The finding is not a confirmed business error.
Configured checks returned 3 inconclusive candidates in this fictional dataset. These are not confirmed business violations. Review each finding against the relevant rule and source evidence.
Present the important information in business language first, with rule details and provenance accessible in a dedicated technical view. An inconclusive candidate must never be presented as a verified error.
Connected source access and live multi-agent assistance are planned, subject to engineering qualification. We label them clearly rather than presenting them as working today.
Bounded CSV, XLSX and supported scalar Parquet subsets are available in the trusted-local foundation; not a hosted service.
S3, PostgreSQL, Snowflake, BigQuery, and URL-based imports are proposed future integrations.
Investigator, Validator, Reconciler, and Reporter have completed limited trusted-local demonstrations. Hosted/public AI is not enabled or verified; broader accuracy is not established.