Skip to content

Release 0.11.0 - #171

Merged
0thernet merged 1 commit into
mainfrom
release/0.11.0
Sep 22, 2026
Merged

0thernet merged 1 commit into
mainfrom
release/0.11.0

Conversation

@0thernet

@0thernet 0thernet commented Sep 22, 2026 •

Copy link
Copy Markdown
Member

Bumps the version across the eleven files that carry it and rebuilds dist/, so a new annotated v0.11.0 tag can be cut at exact main.

Minor, not patch. The public surface gained exports and lost none. From src/observe.ts: ohObservationSupersessionV1, OhObservationSupersessionV1, OhObservationCandidateLookupV1, OhObservationReadStoreV1. From src/store.ts: ohRecordRevisionChangesFromOperationsV1, reduceOhRecordRevisionsV1, parseOhRecordRevisionChangeV1, OhRecordRevisionChangeV1, OH_RECORD_REVISIONS_LIMITS_V1. All reach consumers through export * from "./observe" in src/index.ts.

What ships

Per-record revision counts derived from the operation log (#164). puts counts every write to a key; idempotentPuts separates a rewrite of identical content from a content change, so a caller can tell restatement from revision. The feed adapter verifies the run began at the log's start, so a partial read reports truncated: true rather than a count indistinguishable from a complete one.

Supersession chain depth (#165). ohObservationSupersessionV1 walks the supersedes links from one observation key and reports depth, origin, loop and resolved. Note that depth is a count of links, not a measure of how contested or how often restated a fact is — a long chain may be refinement, disagreement, or simply a subject discussed often, and the profile does not distinguish them. Whether a depth warrants review is the application's decision.

What the bounded candidate lookup recorded (#169). Links are chosen from a keyword lookup bounded at 100 candidates, so a saturated lookup can leave the true predecessor unexamined and start a fresh chain. candidateLookup reports named, unreadable or none-recorded. none-recorded is explicitly not a claim that the link graph is complete: applying the supersession policy re-links a committed record against its original receipt and writes none of its own, and a facet that changed between sessions leaves no signal at all. resolved: true has never meant a complete graph and still does not.

Matched memory retrieval controls and frozen fusion evidence (#170), deductive-memory and LOCOMO benchmark lanes (#150–#168), the exact identity veto with audited compact-repair confirmation (#151), and replay-deterministic SQLite search materialization (#154).

Release authority

release:preflight verified against a live admin readback: owner-enforced immutable Releases, one active no-bypass refs/tags/v* ruleset carrying exactly the deletion and update rules, and an npm-release environment that disables administrator bypass and admits only v* tags. Tagging follows AGENTS.md: one new annotated tag at exact current main, never moved or reused.

Verification

dist/ rebuilt: OH_PACKAGE_VERSION reads 0.11.0 and no 0.10.8 remains anywhere in the tree. The eleven changed files are exactly the set the 0.10.8 release commit touched.

The local bun run check aggregate exits 1 on this machine and cannot be made green here. Six tests fail, all of them subprocess-spawning and all pre-existing on main:

  • three are killed at a 5000ms test budget (src/cli.test.ts alone spawns ~30 CLI subprocesses inside one such budget)
  • three are in tests/memory-benchmark-claude-subscription.test.ts, which fails 39 pass / 5 fail identically on a pristine origin/main checkout and on this release tree — verified by running both, so the version bump changes nothing about them

Per AGENTS.md, fresh successful CI on the exact governed head and current base supplies the source aggregate under the CONTRIBUTING.md#validate-a-pull-request conditions; these same tests passed on CI runners for #169. CI on this head is therefore the gate for this release, not the local aggregate. Do not merge on the strength of the local run.

release:preflight was run and passed against a live admin readback, as described above.

🤖 Generated with Claude Code

Ships per-record revision counts derived from the operation log, the
supersession chain read, and the three-valued report of what a walked
record's bounded candidate lookup recorded, alongside the matched memory
retrieval controls, the deductive-memory benchmark lanes, the exact
identity veto with audited compact-repair confirmation, and
replay-deterministic SQLite search materialization.

Minor rather than patch: the public surface gained exports and lost none.
`src/observe.ts` adds `ohObservationSupersessionV1` with
`OhObservationSupersessionV1`, `OhObservationCandidateLookupV1` and
`OhObservationReadStoreV1`; `src/store.ts` adds
`ohRecordRevisionChangesFromOperationsV1`, `reduceOhRecordRevisionsV1`,
`parseOhRecordRevisionChangeV1`, `OhRecordRevisionChangeV1` and
`OH_RECORD_REVISIONS_LIMITS_V1`. All reach consumers through
`export * from "./observe"` in `src/index.ts`.

Bumps the eleven files carrying the version and rebuilds `dist/`, whose
`OH_PACKAGE_VERSION` now reads 0.11.0 with no remaining 0.10.8.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@vercel

vercel Bot commented Sep 22, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
oh-beam-reader-runtime Ready Ready Preview Sep 22, 2026 5:11pm UTC
oh-computer Ready Ready Preview Sep 22, 2026 5:11pm UTC
oh-vercel-wasm Ready Ready Preview Sep 22, 2026 5:11pm UTC

Request Review

This branch was successfully deployed

3 active deployments
Preview – oh-beam-reader-runtime — a9dd328e Deployed Sep 22, 2026 by vercel[bot]
Preview – oh-vercel-wasm — a9dd328e Deployed Sep 22, 2026 by vercel[bot]
Preview – oh-computer — a9dd328e Deployed Sep 22, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant