Gold star in Rig: mark the tables your AI is endorsed to build on

    A warehouse holds hundreds of tables: raw, staging, experimental, half-duplicated. Gold star is a trust signal at the table-selection step. It's how your data team marks the clean, canonical tables that are safe to build on, so every generated answer starts from the right set.

    How Rig enforces quality

    Gold star is a quality badge you apply to a warehouse table in Context Studio. It is one of four curation layers arranged along the generation pipeline, each answering a different question at its own level.

    Build a revenue retention dashboard by segment
    RBAC
    Entry gate: scopes what every stage can see

    01 · Understanding

    Context

    Whole warehouse

    "What does this mean? Where do I look?"

    Descriptions, business terms, and areas teach the AI what your data means.

    02 · Table selection + publish

    Gold star

    Per table

    "Which sources do we trust?"

    Endorses the vetted, canonical tables that generated work builds on.

    03 · SQL writing

    Metrics

    Per KPI

    "How do we compute this number?"

    Canonical definitions so every KPI is computed the same way.

    Usage rules

    Per table or topic

    "How do I avoid misusing this table?"

    Soft guidance on how a table should and should not be queried.

    Exit gate: re-checks the SQL that actually runs
    Governed dashboard, built on trusted tables
    Diagram: a request to build a revenue retention dashboard flows through Rig's three curation stages, all running inside an RBAC boundary. RBAC first scopes what every stage can see. Stage one, Context, covers understanding across the whole warehouse: what does this mean and where do I look. Stage two, Gold star, covers table selection and publish at the per-table level: which sources do we trust. Stage three is SQL writing, covered by two layers together: Metrics, per KPI, answer how do we compute this number, and Usage rules, per table or topic, answer how do I avoid misusing this table. RBAC then re-checks the generated SQL at execution, and a governed dashboard is returned, built on trusted tables.

    The four layers answer the same big question, how should AI read this data well: context makes the warehouse understandable, gold star narrows generation to trusted sources, and at the SQL writing stage metrics pin down how numbers are computed while usage rules keep individual queries on the rails.

    Mark a table gold star

    Gold-starring a table takes one click in Context Studio:

    1. Open a table in Context Studio.
    2. Click Mark gold star. The button turns amber and reads Gold star, and a gold star badge now appears on that table everywhere it is shown. Click again to un-star.
    3. Use the Gold star filter on the Context page to list only the vetted tables.

    Marking a table requires edit permission, so a data_engineer or admin role. You can also do it in bulk, or hands-free: ask Rig, or an agent connected over MCP, to gold-star a set of tables and it calls the tag_gold_star tool for you, for example "gold-star the marts tables in the revenue area."

    ⭐Gold-star tables are vetted for dashboards and VDMs. That is the promise the badge makes to everyone else on your team, and to every AI tool reading your context layer.

    Discovery by default, enforcement when you are ready

    Gold star has two modes, controlled by a single per-organisation toggle. It ships switched off, so gold star starts as a gentle signal and only becomes a hard rule when your data team decides it is ready.

    Default

    Discovery mode

    Gold star is purely a signal. Vetted tables carry an amber star badge everywhere they appear, and they rank first in table search. Nothing is filtered, so every table stays fully usable. This is the place to start while your team curates the right set.

    Opt in

    Enforcement mode

    Generation is hard-restricted to gold-star tables. Dashboard generation, VDM materialization, and metric hydration draw only on the vetted set. Non-gold tables stay queryable and remain in the context layer for lineage, they just cannot seed new generated work.

    A VDM (Virtual Data Model) is an LLM-enriched or snapshotted table Rig materializes from your warehouse data. You can read more in the Virtual Data Models guide. In enforcement mode, a VDM can only be built on vetted sources.

    Explore the rest of our Context Studio

    Gold star pairs naturally with the rest of Context Studio. Star the canonical tables, set who can see them with access controls, then let Rig build dashboards, models, and metrics on the set you trust.

    Context StudioRole-based access control

    Was this guide helpful?