Skip to content

feature-roadmap.md

Latest commit

 

History

History
154 lines (115 loc) · 6.68 KB

feature-roadmap.md

File metadata and controls

154 lines (115 loc) · 6.68 KB
Raw

Feature roadmap

Last updated: 2026-07-25

This is the ordered delivery list for the next 90 days. It implements the proof-first direction in product-roadmap.md and uses the operating gates in roadmap.md. Earlier FIX, EXP, and RR lists are research history, not parallel queues.

Status

The previous feature list shipped across more than 2,100 published scorecard pages. Daily scorecards, program rollups, alerts, finding-clearance records, provenance, data contracts, the read API, MCP server, badges, and the reusable Action and its Marketplace listing are available. They now support the work below.

The active constraint is not missing dashboard surface. It is the lack of real evidence that an accepted request can be linked to a comparable change in the intended published feed.

Active sequence

Work through this list in order. Do not start later product expansion to fill a participant or release dependency.

1. Publish and verify v1.4.0

Status: completed 2026-07-25. The signed release and the project's first public Marketplace listing are live, and both supported refs passed the downstream consumer check.

Done means:

  • immutable v1.4.0 and signed floating v1 tags point to the intended commit;
  • the release carries the manifest signature, certificate, SBOM, VEX, and provenance assets;
  • the Marketplace listing is publicly visible in Code quality and Testing;
  • clean downstream runs succeed against both tags and produce JSON and HTML.

2. Recruit and contract the pilot

Status: active.

Recruit one support-program liaison and two feed maintainers or vendors. Agree on feed identity, owner role, existing handoff channel, privacy boundary, and what a valid recheck means. Select recurring findings with a concrete export or data change.

Wasco Dial-a-Ride is one evidence-backed recruitment lead because its official Caltrans DDS archive is still the January 2022 upload and scores as expired. Participation and feed identity still require explicit confirmation.

Done means at least two participants accept a named request and the project has six suitable requests available for the 90-day window.

3. Produce the first closure receipt manually

Status: depends on a participant request.

Use shipped provenance and private evidence paths. Preserve before and after bytes, tool versions, feed identity, and the finding fingerprint. Link the accepted action privately. Count a closure only after the intended feed is rechecked under a comparable measurement contract.

Done means an independent reviewer can reproduce the receipt. A finding that merely disappears is still a finding clearance.

4. Remove the smallest repeated friction

Status: evidence-gated.

After the first manual cycle, automate only a step that repeatedly consumes time or creates evidence risk. Candidate work includes a private request manifest, a receipt schema, or one handoff adapter. Do not build a generic workflow UI before the pilot identifies the actual channel.

Done means the change reduces hands-on support time without weakening the identity or comparability checks.

5. Complete six requests and audit the outcome

Status: depends on steps 2 through 4.

Run three cycles, audit every claimed closure, interview participants, and measure time to owner and hands-on support. Apply the pass and stop conditions in roadmap.md.

Done means the project has a documented decision to automate, narrow, change, or stop the workflow. Marketplace installs and scorecard traffic do not decide this gate.

Queue after a passing pilot

This queue is ordered but not active:

  1. Publish the open receipt schema and verifier.
  2. Build the agency-owned quality passport.
  3. Rank pilot notice playbooks by observed evidence.
  4. Productize one existing handoff integration.
  5. Add the procurement acceptance record.
  6. Extend program views with privacy-safe aggregate outcomes.

Each item must remain independently shippable and must preserve the fail-closed receipt contract.

Maintenance lane

Keep daily scoring, security, accessibility, source curation, and release health green. Complete renderer decomposition, registry sharding, queue-backed compute, or hosted scoring only when the measurable triggers in roadmap.md fire.

One direct consumer-feedback slice is also in this lane because it repairs an existing surface without changing a grade or adding a service:

  • Shipped 2026-07-16: primary navigation to the feature finder, translations.txt detection and language filtering, aligned API and CSV fields, a visible coverage limitation, mode-aware language, the ungraded ferry profile, and the auditable European beta gate.
  • Shipped 2026-07-17: the reviewed European depth wave (27 records from the Spain, France, United Kingdom, Germany, and Italy queues; the cohort is now 42 records in 13 countries, with rejections documented), and the interface-localization readiness layer (app string catalog, en-XA pseudolocale preview, right-to-left browser check, hardcoded-string and directional-CSS ratchets; ADR 0038).
  • Shipped 2026-07-25: Mobility Database proposal receipts now account for every recognized Schedule source row with a versioned mechanical disposition ledger. Schema 1.2 validates receipts before either output is written and binds each external identity to the public registry record that currently carries it. Proposals remain separate from human identity, rights, attribution, and admission decisions (ADR 0043).
  • Next only through curation: work the European candidate queue toward the beta gate in global-expansion.md. Feed count alone does not pass source, license, identity, freshness, or country-spread review. Sweden needs Trafiklab API credentials before any record can be reviewed.
  • Still steward-gated: full interface localization (a production language needs a named steward; the engineering prerequisites are in place) and any NeTEx ingestion.

Not on the ship list

  • More general score, map, viewer, catalogue, or validation surfaces.
  • Public agency or vendor rankings.
  • A public raw-feed or continuous cross-agency realtime archive.
  • Broad international expansion without the source, license, identity, and local-steward gates.
  • Consumer-app scraping, multimodal health scoring, or a replacement ticket system.
  • AI-generated repair or automatic closure in the trusted path.

Revisit this file when a pilot cycle finishes or a maintenance trigger fires. Otherwise the order does not change.