Reduce writer contention during vector repair discovery - #206
Open
Coding-Dev-Tools wants to merge 2 commits into
Open
Reduce writer contention during vector repair discovery#206Coding-Dev-Tools wants to merge 2 commits into
Coding-Dev-Tools wants to merge 2 commits into
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ad6f7c6d67
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".
9 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Depends on #203 and targets
codex/rework-memory-foundation. This keeps the existing release candidate unchanged while making repair discovery a separate reviewable change.Finding one erasure behind queued updates previously reserved SQLite's writer for every skipped update. Discovery now reads canonical headers in 100-row pages before acquiring the writer and stops the filtered iterator as soon as its provider-attempt budget is spent, including after failed cleanup.
The memory join uses the Store workspace predicate, retaining excluded records for cleanup and allowed historical records for indexing. Publication still revalidates generation, current canonical state, quarantine and vector identity inside the writer reservation, then acknowledges only the applied generation. Public signatures, return fields, schema and ranking defaults are unchanged. No migration is needed.
Adds real sync/storage regressions, a reproducible repair-only probe with source/driver identities and persisted failed attempts, and documentation of the remaining import/scheduling dependencies. Whole-queue scans, failed-deletion fairness, provider deadlines and full-engine capacity remain separate work.
Type
Verification
ad6f7c6dand fixed indc4382d1. All final focused cases pass.ruff check .passespython -m eval.harness --dataset eval/datasets/sample.jsonl --k 5passesdc4382d1: full-stack Python 3.12 has 5,097 passed / 18 skipped; browser suite has 156 passed; coverage is 85.27%. Python 3.9 floor, full-stack 3.10-3.14, encryption, package, integration and Docker gates passed.Final-source repair probe (
dc4382d1versus31da32c0), five independent processes per variant/backlog:All 20 final runs verified erasure. The same driver uses disposable file-backed SQLite databases, fixed vectors and a synchronous fixture adapter. Timing includes instrumented discovery, writer acquisition, fixture publication and queue counts. Hardware: Intel i7-10700KF, 16 logical CPUs, 34,221,301,760 RAM bytes; Windows 11, CPython 3.12.10, NumPy 2.4.5, SQLite 3.49.1. No model, tokenizer or network call is measured. These are synthetic maintenance results, not the 100k capacity or semantic-quality gate.
Three bounded internal reviewers and the integration owner reviewed the slice; the fourth requested worker could not start because the tool reported its agent-thread limit. Their findings and the GitHub workspace-binding finding were reproduced and fixed. No descendants or separate tasks were used. The addressed GitHub thread was reverified against dc4382d and resolved after the user's explicit authorization.