feat(reference): publish what an artifact CONTAINS, not just that it has a shape - #663
Open
beyondnetPeru wants to merge 2 commits into
Open
feat(reference): publish what an artifact CONTAINS, not just that it has a shape#663beyondnetPeru wants to merge 2 commits into
beyondnetPeru wants to merge 2 commits into
Conversation
…has a shape The registry names an artifact and points at its schema's $id. A consumer therefore learns that a PRD is required in discovery and still cannot find out what a PRD is supposed to contain: an $id is an identity, deliberately not a location, and nothing dereferences it. That gap is not academic. The satellite waiting on it evaluates gate criteria against a flat field map, and with no fields a tenant can configure a rule over a document that nothing will ever read — the gate ends up checking that a file exists and never what it says. Measured on a live tenant before this: 34 catalogued artifacts, zero field schemas with any field, zero criteria. So each registry entry now carries its FIELDS, derived from the schema the Core already ships. The schemas are not rewritten flat: they stay the source and this is a projection, so a schema change propagates on the next read instead of needing a second file kept in sync. Across the corpus that is 529 fields from 50 schemas — the PRD alone goes from nothing to 18, with types and requiredness. The type vocabulary is small on purpose: exactly what the existing criterion operators can judge. A type outside it yields a field no criterion can evaluate, which is worse than a missing one because it can be selected and never satisfied. Collections are omitted for the same reason — gte, in-set and regex all assume a single value — and REPORTED rather than dropped quietly, so someone counting 13 sections against 18 fields can see the difference is arrays and not a truncated schema. An unreadable schema leaves that one artifact without fields instead of failing the registry: one malformed file must not take down the catalogue every other artifact needs. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Signed-off-by: aarroyo <beyondnet.peru@gmail.com>
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
📊 Bilingual Coverage ImpactPR Changes
Repository Coverage
✅ Good: All EN changes have ES counterparts. Generated by GitHub Actions |
The labels these produce become the labels of a FORM — the satellite renders one input per field — so the case matters. Title Case makes a form read like a menu of commands rather than a set of questions, and it clashes with the sentence case the rendering surfaces use everywhere else; two cases on one screen look like two systems sharing it. The acronym list exists because there is no rule to replace it. Lowercasing every word turns `technicalFeasibilityId` into a label ending in "id", which reads as a mistake, and leaving the camel case alone gives "Id", which reads as a typo. Nothing in the spelling separates `id` from `is`, so the terms that get shouted are named one by one. The list is short deliberately: a term missing from it comes out as an ordinary word, which is merely plain, while a term wrongly in it comes out shouting. None of this runs for a schema that publishes a `title`. That is words chosen by whoever owns the shape, and no amount of string-splitting here improves on them. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Signed-off-by: aarroyo <beyondnet.peru@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes the half of the artifact registry a satellite could not use.
The gap
The registry names an artifact and points at its schema's
$id. A consumer therefore learns that a PRD is required in discovery and still cannot find out what a PRD contains — an$idis an identity, deliberately not a location, and nothing dereferences it.That is not academic.
evolith_trackerevaluates gate criteria against a flat field map: a criterion resolves afieldPathand compares it. With no fields there is nothing to resolve, so a tenant can configure a rule over a document and nothing will ever read it — the gate checks that a file exists and never what it says.Measured on a live tenant before this change:
What this does
Each registry entry now carries its fields, derived from the JSON Schema this repo already ships.
The schemas are not rewritten to a flat shape. They stay the source and this is a projection, so a schema change propagates on the next read instead of needing a second file kept in sync.
Across the corpus: 529 fields from 50 schemas. The PRD alone goes from nothing to 18, with types and requiredness:
Three decisions worth reviewing
The type vocabulary is small on purpose —
text,rich-text,number,date,boolean,enum,url. It is exactly what the consumer's criterion operators can judge. A type outside it produces a field no criterion can evaluate, which is worse than a missing one: it can be selected in a rule and never satisfied.Collections are omitted, and reported.
gte,in-setandregexall assume a single value, so a list has no operator that can judge it. They are not dropped in silence:omittedFieldssays which and why, so someone counting 13 sections against 18 fields can see the difference is arrays and not a truncated schema. Corpus-wide that is 248 paths, all declared.An unreadable schema costs one artifact, not the catalogue. A malformed file leaves that entry without fields rather than failing the whole registry, which every other artifact depends on.
Verification
GET /phases/discovery/artifactsreturns the PRD with its 18 fields, and the Tracker'sartifact-field-schemas/prd— which used to answerbaseFields: []— now resolves themOrdering
The satellite side is a separate PR in
evolith_tracker. This one should land first: with the Tracker's change deployed against an older Core, its artifacts simply have no fields again until this follows.🤖 Generated with Claude Code