Repository navigation
docs: clarify token savings evidence and add benchmark protocol - #15
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 READMEs presented 71.5x as session savings and 499x as measured query savings without published reproduction data. This change labels both as historical author-reported claims and explains that ~20,000/~280 is an approximately 71.4x retrieval-context illustration, not an end-to-end session measurement.
Updates English and PT-BR together, preserves the reported inventory with evidence limitations, and adds an audit tracing the claims to April 12 commits. Documents the Hlido date discrepancy and distinguishes repository review from token validation. Related: #12 (does not close the right-of-reply issue).
Adds a paired benchmark protocol covering pinned inputs, task correctness, normal source-navigation baseline, repeated isolated runs, cache accounting, construction/update costs, and separate query/session scopes. Includes an offline normalized-usage summarizer and explicitly synthetic fixture. No private data, paid executions, or new performance claims.
Validation: README parity (12 sections / 37 fences each), JSON examples, Python compilation, diff whitespace, and offline arithmetic/negative-input checks (multiple requests, missing pair, duplicates, negative counts, empty ledger, zero denominators) passed. Full CI/link checks are left to GitHub Actions.
No merge or auto-merge requested.