Skip to content

ROADMAP.md

Latest commit

 

History

History
72 lines (60 loc) · 4.15 KB

ROADMAP.md

File metadata and controls

72 lines (60 loc) · 4.15 KB
Raw

Roadmap

2026-07-18 — Foundation (complete)

  • Local package loader, bounded PDF inspection, report renderers, declarative source-cited rules, and tests.
  • Metric baseline: make verify is the merge gate; branch coverage threshold is 90%; no runtime network calls or retained package data.

Now — Permissioned practitioner pilot (in progress)

  • Done: local controlled-label pilot evidence kit (pilot init and pilot summarize); it stores no package content or free-text rationale.
  • Done: desk review of current official LCI State Clearinghouse guidance; NOD/NOE-specific rules are explicitly gated as experimental. See the source review.
  • Done: participant outreach and evaluation materials that preserve the local, permissioned pilot boundary. See the pilot partner kit.
  • Done: a synthetic-package generator (ceqa-preflight synth) with seedable, objective defects, so reviewer calibration, demos, and false-positive measurement can begin before any real package permission is granted. A committed example lives in examples/.
  • In progress: validate NOE and NOD rule wording with qualified users.
  • Blocked externally: recruit participating organizations, obtain written package permission, and secure two qualified reviewers per activated rule.
  • Measure false-positive rate and reviewer time on synthetic or permissioned packages; publish aggregate results only.
  • Exit criteria: at least two qualified reviewers approve each activated rule; no untriaged high-severity parser or privacy finding.

Before first tagged release

  • Complete an accessibility review of console and HTML outputs with recorded assistive-technology evidence.
  • Implement the internationalization release gate: gettext entry point, reviewed English and Spanish catalogs, catalog parity checks, and deterministic CLI locale selection.
  • Publish a residual-risk review, SBOM, provenance, and release notes.
  • Exit criteria: all CI/security checks green, release checklist signed, no open critical vulnerability, and the release remains advisory-only.

Not planned

  • Hosted document storage, automatic CEQA submission, legal sufficiency determinations, or model-driven legal advice.

Metrics ledger

Last reviewed: 2026-08-07. Owner: maintainer. This is the enforcement ledger required by the portfolio Quality & Metrics standard: each row is an AUTO gate (merge-blocking), a REVIEW gate with a durable evidence artifact, or an explicit N/A with a reason. Values are project-specific; the rigor is defined by the owning standard.

Metric Target Measured by Gate Owner
Branch coverage >= 90% make test (pytest-cov, fail_under = 90) in CI AUTO Maintainer
Lint / format / types 0 errors make lint (ruff), make typecheck (mypy --strict) AUTO Maintainer
Cyclomatic complexity <= 10 per function ruff C90 in make lint AUTO Maintainer
Static security findings 0 unresolved make security (bandit) and CodeQL in CI AUTO Maintainer
Known-vulnerable dependencies 0 in uv.lock make audit (pip-audit) in CI AUTO Maintainer
Secret leaks 0 gitleaks in the security workflow AUTO Maintainer
SHA-pinned workflow uses: 100% portfolio conformance tripwire (weekly) plus PR review REVIEW Maintainer
EN/ES catalog parity 100% keys and placeholders before first public tag i18n release gate; not yet implemented (pre-release gap) REVIEW Maintainer
Accessibility review of console/HTML output Recorded assistive-technology evidence per release Release checklist; no tagged release yet REVIEW Maintainer
Pilot false-positive rate Measured on synthetic or permissioned packages; aggregate results only pilot summarize evidence kit REVIEW Maintainer

AI-DEV-MEASUREMENT: APPLIES. Development is AI-assisted; delivery and quality-debt metrics are measured by portfolio automation from Git and CI. No AI runtime feature ships, so Track B product evaluations are N/A (see the responsible technology audits).