Repository navigation
Qualify native memory and RAG with upstream dashboard evidence - #40
Merged
seathatflowsinourveins merged 3 commits intoSep 21, 2026
Merged
Conversation
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
seathatflowsinourveins
deleted the
codex/memory-rag-alignment-20260921
branch
September 25, 2026 18:47
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The memory service had no semantic embedder, and code search ranked copied retrieval receipts from a nested checkout above the implementation. The native profile now uses upstream local MiniLM; the existing code watcher honors a scoped ignore file, and the refreshed upstream graph describes the actual project.
The portable recipe, catalog and handbook retain native command results, model/source hashes, isolated backup/restore evidence and independent Loki/Grafana observations. The dashboard adapter exposes bounded embedding mode and coverage metadata. Private page bodies, credentials and session identifiers are excluded from public evidence. Exact local report attachments remain downloadable.
Validation: native startup embedded 52 pages with zero failures; the scoped repeat skipped 52 current pages. A real Desktop semantic query returned the shared-memory decision where CLI text search returned no hits. All 52 restored latest-page bodies matched the immutable backup. Qdrant's complete 196-point scan excluded copied checkout paths; the same code query ranked the implementation first. Native Loki and browser values agreed. All 60 report attachments passed browser-side hash checks and a downloaded native status matched its source. The project suite passed 652 tests with 41 skips; 24 adapter checks, catalog/integrity/convergence checks and generated HTML validation passed. Independent review findings were resolved.
These are bounded native integration results. Broad retrieval benchmarks, fresh Claude semantic queries, additional graph controls and Qdrant recovery remain unqualified. Memory browser search remains FTS5 and LLM consolidation stays disabled. Stored vectors can include superseded page versions; coverage is reported separately. The compatible vLLM 0.25.0 pin is retained after the previously evidenced 0.29.0 WSL initialization failure.