Nova Codex Curator

Questions or comments?

Post them as a comment on the tracking issue -- requires a GitHub account with access to this repo.

AI-DRAFT. Seeded from the Nova Codex Curator PRD v1.5 (Vanessa Gutierrez, PM; Ravindra Kumar, Lead Developer; July 9, 2026). Not yet reviewed by PM/EL against this repo's schema-first gate or Feature-Set conventions — treat scoping and the schema delta below as a starting point to react to, not a committed plan. Source doc kept at reference/Nova_Codex_Curator_PRD_v1.5.docx for durability.

Business case

Nova Policy Advisor's compliance answers are only as trustworthy as the regulatory content grounding them. Today that content covers U.S. federal regulations only — state-level content (starting with California and New York) and Canadian federal regulations have been scoped and sourced but not yet ingested, and there's a backlog item to crawl Canadian regulations specifically for Ateko, an active customer waiting on that coverage.

The engineering team responsible for ingestion has no structured, repeatable pipeline or review workflow for adding a new jurisdiction — ingestion has been ad hoc, and there's no internal tool to see, at a glance, what's covered, what's pending review, and what's missing entirely. Every HR/ Compliance Manager relying on Nova Policy Advisor is exposed if the underlying corpus is incomplete or unaudited, and the advisor can't yet hyperlink its answers back to the specific regulation text it's citing — a Curator-side data-completeness gap as much as an Advisor-side UI gap.

The measurable cost: coverage expansion is bottlenecked on ad hoc engineering effort rather than a scoped, staffable process; Ateko is waiting on Canada coverage specifically; and without a shared way to see curation status per country, product and engineering have no shared source of truth for planning. Roadmap: prioritize a couple of additional states in Q3 (building on the CA/NY source lists already gathered), add more in Q4, with Canada prioritized ahead of that because of Ateko's active need.

Codex Curator is internal infrastructure, not an employee-facing or customer-facing product — customers never see it directly, though they feel its coverage gaps through Nova Policy Advisor.

Goal

Give the Curator engineering team and SME reviewers a defined, repeatable pipeline — discover candidate sources → SME curates the source list → crawl → SME reviews individual documents (accept / reject / needs review) → sync to the vector DB — with a dashboard showing exactly where every country stands. Customer-recognizable outcome (indirect, since Curator itself is internal-only): standing up a new jurisdiction, like Canada for Ateko, goes from ad hoc engineering effort to a scoped, staffable, auditable process — and Nova Policy Advisor's answers for that jurisdiction become progressively better-grounded (Tier 3 → Tier 2 → Tier 1) as that process runs, rather than jumping straight from "no coverage" to "fully curated" with no visibility in between.

Nothing reaches the vector database without passing two human gates: a candidate source being added to (or kept off) a country's approved source list, and each individual crawled document being accepted, rejected, or flagged for review. This Feature Set does not propose relaxing either gate.