Skip to content

Changelog

Things we shipped recently. The most recent first.

Files, onboarding & account security

  • New engine: query files in object storage — Parquet, Delta Lake, CSV, JSON, and Excel on S3, Azure, GCS, or HTTPS — powered by DuckDB.
  • Guided onboarding: connect, discover, and run your first query on a sample dataset in minutes.
  • Redesigned sign-in page, plus password reset and email verification for email accounts.
  • Long-running schema analysis is now a background job — close the tab, come back, resume or retry without losing progress.

Discovery at warehouse scale

  • Multi-schema discovery with per-schema checkpoints and partial-failure recovery.
  • Google sign-in can auto-connect BigQuery — one consent, zero credential pasting.
  • Row-level security backstop on every user-data table, on top of per-request ownership checks.
  • Schema & Discovery page: edit descriptions, key tables, and relationships that feed the planner.

Charts, dashboards & palette v2

  • Charts are now spec-driven Vega-Lite compiled by the orchestrator — deterministic, restyleable, exportable as PNG/SVG.
  • Dashboards refresh from stored SQL without an LLM in the loop; conversational ops like "add this to my Sales dashboard".
  • Chart style editor with theme-aware chrome (app, ocean, dark, custom).
  • Brand palette v2: deeper ultramarine accent, refreshed ocean and dark themes.

Brand · datarelix.ai

  • Renamed the product from EOADC to datarelix.ai across the app, marketing, and docs sites.
  • Refreshed marketing site: features, pricing, security, contact, privacy, terms.
  • Improved dialog backdrop legibility on dark themes.
  • New conversation search, connection health indicators, table horizontal scroll affordance.

Athena & cross-dialect parity

  • Added Amazon Athena support across the dialect stack (MCP, introspection, planner, UI).
  • Canonical dialect registry — one source of truth for IR validation, planner prompts, MCP routing.
  • Better scope normalization for BigQuery, Athena, and Kusto.

Trust calibration

  • EvidenceRow on every assistant message — tables used, row count, runtime.
  • Error taxonomy with contextual recovery: NO_RESULTS, TIMEOUT, PLAN_FAILURE, CONNECTION_FAILURE.
  • Onboarding rail in chat: API key → connection → discovery → first query.