Hypercerts description & workScope union fix

pi-simocracy · fixed

src/writes.ts · createProposal assigned bare strings to record.description and record.workScope, and forwarded contributor identities as bare strings. lex-gql silently rejects these, so any proposal authored from pi never made it to the indexer (proposals appeared on the user's PDS but were absent from simocracy.org/proposals).

// before — silently dropped on the way to the indexer
record.description = body;
record.workScope   = ws;

// after
record.description = { $type: "org.hypercerts.defs#descriptionString", value: body };
record.workScope   = { $type: "org.hypercerts.claim.activity#workScopeString", scope: ws };
record.contributors = opts.contributors.map((c) => ({
  contributorIdentity: {
    $type: "org.hypercerts.claim.activity#contributorIdentity",
    identity: c.contributorIdentity,
  },
}));

simocracy-v2 · fixed across two write paths + the read pipeline

The webapp has two proposal-form components and they disagreed with each other:

• components/ftc-sf/proposal-form-dialog.tsx wrote bare strings (same bug as pi-simocracy) for description, workScope, and contributor identities.
• components/events/gathering-proposal-form.tsx built a Leaflet document for description but forgot the top-level $type: "pub.leaflet.pages.linearDocument", so the lex validator couldn't tell which union variant the value was.

On the read side, HypercertActivity.description was typed as string in lib/lexicon-types.ts — a half-truth that worked only because the indexer happened to flatten descriptionString back to a scalar. Anyone calling .slice() on it (e.g. generateMetadata on the proposal permalink page) would crash the moment a real Leaflet doc came through.

Fix:

1. Both forms wrap their union writes with $type (and the Leaflet form now sets the top-level discriminator).
2. HypercertActivity.description is now typed as ActivityDescription = string | DescriptionString | LeafletDescription | StrongRef.
3. descriptionToText() in lib/budget-items.ts now handles all four shapes (string, descriptionString, Leaflet doc, legacy {text}).
4. Every direct consumer (.slice, edit-form pre-fill, S-Process LLM payload) was wrapped through the normaliser.

// lib/budget-items.ts — single source of truth, used at every read site
export function descriptionToText(input: unknown): string {
  if (input == null) return ""
  if (typeof input === "string") return input          // indexer-flattened
  if (typeof input === "object") {
    const obj = input as Record<string, unknown>
    if (typeof obj.value === "string") return obj.value           // descriptionString
    if (Array.isArray(obj.blocks)) {                              // leaflet doc
      return obj.blocks
        .map(e => (e as any)?.block?.plaintext)
        .filter((s: unknown): s is string => typeof s === "string")
        .join("\n\n")
    }
    if (typeof obj.text === "string") return obj.text             // legacy
  }
  return ""
}

Fixed — pi-simocracy and simocracy-v2 silently incompatible writes

Both were emitting { description: "...", workScope: "..." } as plain strings. The records land on the user's PDS but never propagate to the Hyperindexer, so they don't appear in simocracy.org/proposals. After the fix, all three projects share one wire shape that matches the lexicon.

Fixed — read-side type gap in simocracy-v2

HypercertActivity.description was typed string, masking the fact that future indexer revisions could surface raw union shapes. Retyped to ActivityDescription and wrapped every direct consumer (.slice, edit-form pre-fill, LLM evaluator payload) through descriptionToText() so the read pipeline now tolerates all four shapes.