Skip to content

Releases: swimit-io/whyve

Release list

Whyve 0.3.0

Choose a tag to compare

@Jeis-Jw Jeis-Jw released this 29 Sep 05:51

Breaking. The product version restarts at 0.3.0; the protocol is
context-common/v3 (record schemas context-<kind>/v2, indexes
context-area-index/v2 and context-root-index/v2). See the migration guide.

  • A new normative record model (docs/record-model.md),
    from which the runtime model is generated: flat compact-JSON headers, a common
    key, fixed sections with Korean aliases, a final ## Sources list, namespaced
    host headers (howse.thread), YYYY-MM-DD-title.md names (DOCUMENT and TERM:
    title.md), and plain Markdown index rows.
  • New host API WhyveHost: capabilities, status, list, searchHeaders,
    read, checkSlot, compare, validateReadReceipt, prepare, apply.
    Writes are prepare → apply(handle); prepare returns prepared,
    needs_approval, needs_review, or needs_archive, and apply is idempotent.
  • Meaning is judged by the caller: compare delivers the mandatory records
    (supersede target, slot occupants, typed references) with complete bodies and a
    comparison receipt; prepare requires a same|separate|support|conflict|replace|unclear
    judgment for each. A capture with no mandatory records needs no comparison.
  • Missing evidence, rationale, alternatives, success criteria, or summary are
    quality flags, never rejections. Bodies over 256 KiB return needs_archive
    with suggested ARCHIVE chunks.
  • Discovery is exact: list filters with coverage and cursors; searchHeaders
    reads header blocks with eq/contains/regex conditions. Lexical recall and
    scoring are removed.
  • Authorization is unchanged in spirit: user or configured policy for every
    write. In explicit mode, policy authorization returns needs_approval;
    complete it with prepare({preparedHandle, authorization}).
  • Removed: candidates, attestations, assertions and evidence pointers, same-claim
    digests, approval digests and receipt files, *_workflow.mjs, recall,
    search, checkDecision, compareDecision, specView, revisitDecisions,
    owner routing, and area registration.
  • context-common/v2 vaults are readable; writing requires
    whyve migrate-project PATH --to-format context-common/v3, with a dry run,
    --apply-plan (backup, journaled apply, verification), and --rollback-plan.
  • Skills, rules, templates, managed guidance, and docs describe the new flow with
    JSON input files; the shipped decision examples are executed by the tests.