docs: restore the classification matrix and re-cut the graph for mobile - #862
Merged
Conversation
The README graph dropped the conductor's classification matrix when it was sized down for the column, which left the two conductors looking like an arbitrary hand-off. It is back, centred over the level-2 conductor: the sweep lands on UNCERTAIN, the escalation fires, and level 3 resolves it to CRITICAL. The REJECTED label sat at x=392 while the rejection loop runs between x=586 and x=958, so it read as unattached text. It now sits under the loop's low point. Re-encoded while the frames were being regenerated. The previous encode carried an alpha channel on 396 of 536 frames even though every source frame is opaque RGB, and wrote a full keyframe every ~5 frames. Animated WebP is CPU-decoded with no frame-dropping, so that cost showed up on phones as an animation that ran slow rather than one that stuttered. Opaque frames, sparse keyframes and 15fps cut it from 252 MP decoded per loop to 76 MP, and the pair from 6.6 MB to 1.3 MB. Same 1856px width, so the retina supply ratio holds. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Greptile SummaryThe PR refreshes the README architecture presentation.
Confidence Score: 5/5The PR appears safe to merge. No blocking failure remains.
|
| Filename | Overview |
|---|---|
| README.md | Updates the architecture diagram’s alternative text and caption to match the restored classification and escalation narrative. |
| docs/assets/zeroshot-architecture-dark.webp | Replaces the dark-mode animated architecture diagram. |
| docs/assets/zeroshot-architecture-light.webp | Replaces the light-mode animated architecture diagram. |
Reviews (3): Last reviewed commit: "docs: put the graph back at the native 3..." | Re-trigger Greptile
The bar spanned x 32..1022 while the rule directly above it spans 32..1088, so it read as stopping 66px short -- the gap was the percentage readout parked inline beside it. The bar now runs margin to margin and the readout sits above its right end, flush with the same margin. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The 15fps cut in the mobile re-encode was the wrong lever. Decode cost tracks total changed pixel area, not frame count -- slicing time finer makes each delta proportionally smaller, and identical frames dedup away. Measured on the same source: 15fps costs 75 MP per loop, 30fps costs 91 MP. A 21% increase, not the 2x a frame count would suggest. The mobile win came from dropping the alpha channel and the keyframe every ~5 frames, both of which stay. 30fps still lands 64% under the 252 MP that was choking phones. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
🎉 This PR is included in version 6.17.0 🎉 The release is available on: Your semantic-release bot 📦🚀 |
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.
Three fixes to the README architecture graph.
The classification matrix is back. It was dropped when the graph was sized down for the README column, on the grounds that its 10px axis labels fall under the type floor. That was the wrong call — without it the two conductors read as an arbitrary hand-off, and the matrix registers as a shape (a scan sweep, one rust cell) long before anyone reads the words. It now sits centred over the level-2 conductor: the sweep lands on UNCERTAIN, the escalation fires, level 3 resolves it to CRITICAL.
The REJECTED label was orphaned. It sat at x=392 while the rejection loop runs between x=586 and x=958 — floating text with nothing attached to it. Moved under the loop's low point, and the loop was raised slightly to open the gap.
The animation was slow on phones. Not a rendering bug — an encoding one. Animated WebP is CPU-decoded with no frame-dropping, so when a phone falls behind it stretches the animation instead of skipping ahead. The old encode made that worse two ways: it carried an alpha channel on 396 of 536 frames despite every source frame being opaque RGB, and wrote a full keyframe every ~5 frames.
Width stays 1856px, so the retina supply ratio for an ~890px column is unchanged.
🤖 Generated with Claude Code