Staging - #445
Merged
Merged
Staging#445
Conversation
GitHub dark mode painted every figure on a white canvas: the banner, the parsing and retrieval diagrams, and the benchmark grid. Each figure now ships a light and a dark file selected by a <picture> element with prefers-color-scheme, so the switch needs no CSS. - Banner: the light and dark hero exports replace knowhere-banner.png. - Parsing and retrieval: refreshed from the new 01/02 designs. Those designs ship light only, so the two diagrams stay single-variant for now. - Benchmark: re-rendered from the design SVG at 1600x1120 with a 2x device scale. The dark variant comes from the same SVG with color literals remapped and geometry untouched, holding each element relative contrast (bars 1.56:1 to 1.89:1, grid 1.30:1 to 1.56:1, text at or above 4.5:1). - Sources: both benchmark SVGs live in docs/assets/source/ with the palette mapping and the render recipe. - Dropped the four figures the README no longer references. Text in the benchmark SVG is exported as glyph paths, so future metric changes belong in the generating script, not in the SVG.
The banner sat at the left edge while the headline, badges, links, and every figure are centered, because it was the only image in the file without a centering wrapper. Wrap it in the same <p align="center"> the other figures use.
The new hero banner already renders "Prepare unstructured data for AI agents", so the h1 directly below it repeated the same sentence inside the first screen. Drop the h1 and keep the wording in the banner alt text, where screen readers and search still pick it up.
docs: give the README figures a dark counterpart
GitHub markdown sets max-width: 100% on images but never height: auto, so an explicit height attribute keeps the box at that height while the width shrinks to the README column. The banner was the only image in the file carrying one, so it rendered 768x500 in the column instead of 768x384: the artwork and its headline came out 30% taller than designed. Reproduced the rule locally at a 768px column: width+height renders 768x500 (ratio 1.536 against the natural 2.000), width alone renders 768x384 (ratio 2.000). A template match against the artwork on the live page agreed with the stretched case (0.48 horizontal scale, 0.58 vertical). Dropping the attribute makes the banner behave like every other figure in the file, which set width only.
docs: stop stretching the README banner
Both variants come from the new 3200x1600 exports (2x of the previous 1600x800) with the same composition: lockup top left, headline bottom left, illustration right. The wrappers keep width-only sizing so the aspect comes from the file.
The Vision + Text and Step 2 sections were the last light-only images in the README. Both now ship a dark export next to the light one and switch through <picture>, so every figure in the file follows the reader color scheme. The two dark exports are mapped by content, not by export order: the parsing figure carries the two tracks and the unified schema, the retrieval figure carries the agent list and grounded evidence.
docs: refresh the README banner artwork and add the dark diagrams
…page text handling - Consolidated VLM model references to use `deepseek-flash` across documentation and configuration files. - Enhanced the `PageTextBands` structure to support ordered line records, improving text extraction and processing. - Removed deprecated fields and streamlined the handling of page text in various components, ensuring consistency in data representation. - Updated tests to reflect changes in the `PageTextBands` structure and VLM model usage.
The previous dump flattened every extracted TOC into one tree, which hid later regions in multi-TOC documents. Co-authored-by: Cursor <cursoragent@cursor.com>
…-text refactor: unify VLM on deepseek-flash and tighten page-text handling
…zation-under-10s perf(publication): reduce materialization persistence latency
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
CodeQL found more than 20 potential problems in the proposed changes. Check the Files changed tab for more details.
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
3 participants
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.
No description provided.