Skip to content

MnemoSync architecture and delivery strategy (docs) + Node engines range - #2

Open
ehitimum wants to merge 3 commits into
feature/ai-assistantfrom
codex/mnemosync-foundation
Open

ehitimum wants to merge 3 commits into
feature/ai-assistantfrom
codex/mnemosync-foundation

Conversation

@ehitimum

@ehitimum ehitimum commented Nov 12, 2026 •

Copy link
Copy Markdown
Owner

Establish the implementation contract for optional local-first project sync, starting from the current AI feature HEAD (0f15939) while PR #1 is open. Preserve the supplied development plan and handoff memory under docs/mnemosync/.

Records storage, revision, trust, Drive, merge and recovery ADRs; archive schema and limits; threat model; phased PR strategy; and a verification matrix. Updates architecture, roadmap and development log. Dependent PR #3 implements the native Phase 0 recovery slice. This work does not enable synchronization.

CI exposed two inherited reliability problems. Node 20 cannot start locked jsdom 30 / undici 8; pin Node 24.15.0 for CI and releases, declare the engine and add a fresh-worker DOM regression. Once tests could start, fast Linux runs exposed AI rewrites joining adjacent typing in undo history. Add explicit history boundaries for word rewrites, structural rewrites and continuations, with three fixed-clock regressions preserving prior and subsequent typing. This branch passes npm run check (120 tests, lint, types, build) and 31 regressions locally. The dependent backup implementation also passes its full local checks. A saved permission policy blocks browser visual access; installed Windows and real Android acceptance remain pending.

Targets feature/ai-assistant to preserve PR #1; retarget after that prerequisite merges.

@ehitimum ehitimum changed the title MnemoSync: architecture, threat model and phased development strategy MnemoSync: development strategy and compatible CI runtime Nov 12, 2026
@ehitimum ehitimum changed the title MnemoSync: development strategy and compatible CI runtime MnemoSync prerequisites: strategy, CI runtime and safe AI undo Nov 12, 2026
ehitimum and others added 3 commits November 15, 2026 16:37
- engines.node widened to ^22.22.2 || ^24.15.0 || >=26 (what locked jsdom and
  undici accept; the suite passes on Node 22.23). CI keeps 24.15.0.
- The runtime regression could never fail the way it described (Vitest dies
  before collecting on Node 20); it now asserts the running Node satisfies
  engines.node and that the range refuses Node 20.
- docs/mnemosync/MEMORY.md says at the top that it is kept as working notes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
import.meta.url is not a file URL under jsdom.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@ehitimum
ehitimum force-pushed the codex/mnemosync-foundation branch from 8b4977b to 88a04de Compare November 15, 2026 10:48
@ehitimum

Copy link
Copy Markdown
Owner Author

Changes for the review

This branch was rebased onto the updated feature/ai-assistant (now 2.1.1). The two duplicated commits (2fb06b5, 8b4977b) were checked to have the same code as bdaab93 / 991a20e, and dropped. This PR now contains only the MnemoSync docs and the fixes below.

# Finding Change
1 Duplicate commits Removed by the rebase. Retitled to describe what remains.
2 engines pinned to 24.x "node": "^22.22.2 || ^24.15.0 || >=26" (what locked jsdom/undici accept). CI keeps 24.15.0 from .node-version. README updated.
3 Runtime test could not fail as described Replaced: runtime.regression.test.ts asserts the running Node satisfies engines.node, and that the range accepts 22.23 and refuses 20 and 24.14.
4 Extra history dispatches (note) No change, as suggested.
Docs docs/mnemosync/MEMORY.md A header now says it is kept on purpose as the MnemoSync working notes, not end-user docs.

PRs #3 and #4 were rebased on top of this branch the same way.

🤖 Generated with Claude Code

@ehitimum ehitimum changed the title MnemoSync prerequisites: strategy, CI runtime and safe AI undo MnemoSync architecture and delivery strategy (docs) + Node engines range Nov 15, 2026
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