fix(core): accept-time publication marker; PUT renames announce moves; real-DB runner tests by phernandez · Pull Request #1668 · basicmachines-co/basic-memory

phernandez · 2026-10-07T14:52:45Z

Summary

Closes #1661.

Publication marker in the accept transaction

The accept transaction writes the pending relation_search_refresh marker for the accepted generation, through RelationRepository.record_pending_relation_publication (the single writer, also used by the publisher's begin). An accepted note is never visible without it.
begin reuses a marker already pending for its generation, so a successful publication leaves exactly one unit of refresh work. A failed publication leaves it pending, which keeps change detection's repair path.
Why: Basic Memory Cloud wrote its own marker after the request so its Wiki report would wait. That marker landed after publication cleanup, stayed pending, and masked the entity checksum, so every app-written note was fully re-indexed when its storage notification arrived. Cloud drops that fence once this ships.

PUT renames announce the move

A PUT whose destination differs from the note's path recorded a move on its project change, but its materialization carried no previous path. The materialization now takes previous_file_path from the project change in attach_accepted_project_note_change, so a PUT rename and an explicit move carry the same evidence. The planners no longer accept an independent second value.

Tests

The accepted-note runner tests drove every mutation with a stand-in session and fake repositories. They now run the production mutation dependencies against a real database (test-int/test_accepted_note_mutation_runner_db.py): create, update, edit, move, delete, base-checksum conflicts, relay supersession, materialization states, directory casing, graph policy, pending markers, inbound-link cleanup, PUT-rename move evidence. Races are injected at the note lock and say so. Only pure mapping tests remain as unit tests. Net −2,100 lines.
test-int/test_note_write_outcomes.py: a published write leaves nothing pending; the accepted generation is pending before post-commit publication starts.
Two SQLite engine tests now pass an explicit SQLite config instead of reading the developer's home config.
Each new guard was checked by breaking the fix and watching the test fail.

🤖 Generated with Claude Code

https://claude.ai/code/session_01T8wjd6HrtSA2LN9ssC4NzF

…action An accepted note generation is now never visible without its pending publication marker. The marker used to be written by the post-commit publisher in its own transaction, so a reader between the accept commit and that step saw an accepted generation with no marker, and a hosted runtime that needed one had to write its own after the fact. The publisher's begin step reuses a marker already pending for its generation instead of adding a second one, so a successful publication leaves exactly one unit of refresh work. A failed publication still leaves the marker pending, which keeps change detection's repair path. Closes #1661 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01T8wjd6HrtSA2LN9ssC4NzF Signed-off-by: phernandez <paul@basicmachines.co>

The create-runner tests drove the runner with a stand-in session and fake repositories, so they could not see what the accept transaction writes. They now run the production mutation dependencies in a real transaction and assert the persisted note, project change, graph publication, pending marker, path conflicts and directory casing from the database. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01T8wjd6HrtSA2LN9ssC4NzF Signed-off-by: phernandez <paul@basicmachines.co>

chatgpt-codex-connector · 2026-10-07T14:52:57Z

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-07T16:02:54.591990Z	`6f0735f`	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 37a3348d82

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-10-07T14:58:42Z

+    is accepted; the post-commit publisher converts it once the graph is written, and a
+    publication that fails leaves it pending for change detection to repair.
+    """
+    session.add(


Update the fake sessions for the new marker write

When the unit suite exercises persist_accepted_note_snapshot() or persist_accepted_note_move(), four existing tests pass either a plain object() or _FlushSession, neither of which implements add(), so this new call raises AttributeError before their assertions run. uv run pytest -q tests/indexing/test_accepted_note_write_runner.py now reports 4 failures; route this persistence through the repository protocol or update those fake sessions so the required unit suite remains green.

Useful? React with 👍 / 👎.

Fixed in 4afe067. The marker now goes through the repository protocol: RelationRepository.record_pending_relation_publication is the single writer, used by the accept transaction (via relation_repository(project_id)) and the publisher's begin step. The fake-session runner tests are replaced by real-DB tests in test-int/test_accepted_note_mutation_runner_db.py rather than teaching the fakes the new method.

…ository The accept transaction recorded its pending marker with a bare session.add, the only accept-time write that bypassed the repository capabilities. The marker now has one writer, RelationRepository.record_pending_relation_publication, used by both the accept transaction and the publisher's begin step. The accepted-note runner tests drove every mutation with a stand-in session and fake repositories, so they could not see what a transaction actually writes. They now run the production mutation dependencies against a real database: create, update, edit, move and delete, including base-checksum conflicts, relay supersession, materialization states, directory casing, graph policy, pending markers and inbound-link cleanup. Races are injected at the note lock and say so. Only pure mapping tests remain as unit tests. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01T8wjd6HrtSA2LN9ssC4NzF Signed-off-by: phernandez <paul@basicmachines.co>

A PUT whose destination differs from the note's path records the change as a move, but its materialization carried no previous path, so runtimes that announce moves from the materialization could only report an update and left readers holding the old path. The project change already records what the mutation did; the materialization now takes its previous path from it, so a PUT rename and an explicit move carry the same move evidence and the planners no longer accept a second, independent value. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01T8wjd6HrtSA2LN9ssC4NzF Signed-off-by: phernandez <paul@basicmachines.co>

Both tests built their engine without a config, so the database backend came from the developer's home config. On a machine configured for Postgres they ran SQLite-only SQL against Postgres and failed. They now pass an explicit SQLite config. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01T8wjd6HrtSA2LN9ssC4NzF Signed-off-by: phernandez <paul@basicmachines.co>

…s match Text mode on Windows writes CRLF, so a seeded source object no longer hashed to the accepted Markdown and the move-cleanup cases failed only there. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01T8wjd6HrtSA2LN9ssC4NzF Signed-off-by: phernandez <paul@basicmachines.co>

phernandez and others added 2 commits October 7, 2026 09:45

chatgpt-codex-connector Bot reviewed Oct 7, 2026

View reviewed changes

phernandez and others added 4 commits October 7, 2026 10:19

phernandez changed the title ~~fix(core): record the relation publication marker in the accept transaction~~ fix(core): accept-time publication marker; PUT renames announce moves; real-DB runner tests Oct 7, 2026

phernandez merged commit 23b1b4a into main Oct 7, 2026
51 of 52 checks passed

phernandez deleted the accepted-write-publication-marker branch October 7, 2026 17:09

phernandez mentioned this pull request Oct 8, 2026

docs(core): bring the v0.24.0 changelog up to date #1680

Merged

Navigation Menu

Repository navigation

fix(core): accept-time publication marker; PUT renames announce moves; real-DB runner tests - #1668

fix(core): accept-time publication marker; PUT renames announce moves; real-DB runner tests#1668
phernandez merged 6 commits into
mainfrom
accepted-write-publication-marker

phernandez commented Oct 7, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot commented Oct 7, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Oct 7, 2026

Uh oh!

phernandez Oct 7, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

phernandez commented Oct 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Uh oh!

chatgpt-codex-connector Bot commented Oct 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codex Review Summary

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Oct 7, 2026

Choose a reason for hiding this comment

Uh oh!

phernandez Oct 7, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

phernandez commented Oct 7, 2026 •

edited

Loading

chatgpt-codex-connector Bot commented Oct 7, 2026 •

edited

Loading