docs(readme): lead 'What it does' with the v3.0 four-lane retrieval stack - #729
Merged
Conversation
The v3.0 release-prep PR (b3022b2) parked the four-lane reality behind an apology paragraph because the pre-push discretion grep flagged any rewrite of the introductory sentence (it contains 'Claude Code' and git's line-level diff put the unchanged phrase on a '+' line under a within-line edit). Now that v3.0 is shipped I can rewrite cleanly: drop 'in Claude Code' from the lead sentence (the same phrase still appears elsewhere in the README; it doesn't need to live on this one line), flip 'two-layer search' to 'four retrieval lanes in parallel and merges the result', expand the code block to L0/L1/L2/L2.5 with concrete lane descriptions (BM25 + posterior-weighted rerank, typed-edge BFS over named edge types, Plate-FFT structural bind/probe), and replace the apology paragraph with a one-liner that names the trim policy, the 2,400-token default budget, and the #718 stack-r1-r3 default flip with its bench numbers (+0.2851 NDCG@k absolute / +94.8% / p99 4.5 ms). ARCHITECTURE § Retrieval cross-link preserved.
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Sorry @robotrocketscience, you have reached your weekly rate limit of 500000 diff characters.
Please try again later or upgrade to continue using Sourcery
|
Caution Review failedPull request was closed or merged during review No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
📝 WalkthroughWalkthroughThis PR updates README documentation to replace the prior two-layer retrieval description with the v3.0 default four-lane pipeline (L0 locked beliefs; L1 FTS5 BM25; L2 typed-edge graph walk; L2.5 structural HRR), documents budget-trimming and locked-belief overflow behavior, default 2,400-token budget, and ChangesRetrieval Pipeline Documentation Update
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes Possibly related PRs
Suggested labels
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Two problems with the line as written: 1. It assumes the reader's default is a vector DB and frames aelfrice as giving something up in exchange. 'What you get for free' is the value-prop section; a one-line antifeature in the middle of it interrupts the pitch and answers a question new readers haven't asked yet. 2. It overstates the gap. The v1.7 HRR structural-query lane (default- on as of v2.1, #154) closed the vocabulary-gap-recovery case that embedding pitches usually rest on -- the #433 vocabulary-bridge campaign reframed 'vocabulary bridge' as 'typed-edge structural retrieval' and closed wontfix on that basis. 'No fuzzy semantic recall' is true at the embedding level but misleading at the capability level. The full nuance (free-form paraphrase / synonymy lives in the consuming agent per the v3.0 #605 ratification; embedding systems still beat aelfrice on multi-session aggregation) lives in PHILOSOPHY § Determinism and LIMITATIONS § No semantic similarity. Both are linked one section down in the README. The compressed one-liner here is noise, not signal.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@README.md`:
- Around line 60-65: The unlabeled fenced code block containing the lines
starting with "L0: locked beliefs" (the four-line pseudo-format block) triggers
markdownlint MD040; add a language tag by changing the opening fence from ``` to
```text and keep the closing fence unchanged so the block becomes a text-fenced
code block (ensure the exact snippet with "L0: locked beliefs", "L1: FTS5
keyword", "L2: graph walk", "L2.5: structural HRR" is wrapped by ```text ...
```).
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
robotrocketscience
added
the
ready-to-merge
|
merge-train: blocked 1 review thread(s) are unresolved on these files: README.md. Resolve them on the PR (click 'Resolve conversation' on each) and re-add the label. The |
github-actions
Bot
removed
the
ready-to-merge
CodeRabbit flagged the new fenced block on PR #729 line 60 as missing a language tag (markdownlint MD040). The block carries pseudo-format output (lane name + arrow + description), so `text` is the right tag. Closes the unresolved review thread blocking merge-train per the #721 conversation-resolution pre-check.
robotrocketscience
added
the
ready-to-merge
github-actions
Bot
removed
the
ready-to-merge
|
merge-train: merged bc11469 → |
1 similar comment
|
merge-train: merged bc11469 → |
4 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
1 participant
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.
Summary
stack-r1-r3default-flip bench evidence inline (+0.2851 NDCG@k absolute / +94.8% / p99 4.5 ms) so the README reflects what's actually running onmainafter feat(retrieval): flip query_strategy default legacy-bm25 → stack-r1-r3 (#291 PR-3) #718.Why a follow-up instead of inline in #728
PR #728 wanted to do this rewrite but the introductory sentence contained "Claude Code" and git's line-level diff put it on a
+line whenever the in-line text was edited. The pre-push discretion grep blocks+lines containing that phrase. Workarounds tried in #728:Now that v3.0 is cut, this PR takes option 1: the lead sentence becomes "When you submit a prompt, aelfrice's
UserPromptSubmithook fires before the model sees your message." — "Claude Code" is still mentioned in five other places in the README, just not on this one line.Test plan
+lines contain banned vocabulary.wc -l README.mddelta is +2 (one new lane line × 2 lanes, minus one apology paragraph that collapsed to a one-liner; net +2).Summary by CodeRabbit