Repository navigation
docs: use the annotated sweep chart in the showcase - #520
Conversation
Replaces the plain matplotlib output with a chart that states its own finding: recall@k is set by top_k alone, latency stays flat across the grid, and the best cell is marked. Its footnote also explains why two cells are means, which the earlier chart left unexplained. Padded to the same aspect as the comparison board beside it, so the two cells in that table still end at the same height. Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
gloryfromca
left a comment
gloryfromca
left a comment
There was a problem hiding this comment.
No blockers; this can merge as far as I am concerned.
Reviewed the complete github/main...HEAD diff, surrounding README showcase markup, the introducing history, PR metadata, and the applicable AGENTS.md/CLAUDE.md asset, language, branch, and commit rules. CONTEXT-MAP.md confirms no Runtime/TUI domain vocabulary or architecture boundary is involved; this README-only asset swap has no callers or compatibility surface, and no tests were weakened.
I fetched and inspected both chart versions and the neighboring comparison board. The new attachment loads successfully, is a 2000x1332 PNG matching the old chart and adjacent board, and its annotations support the revised alt text. Local git diff --check, large-file, source-language, and commit-message gates pass. GitHub's docs/page/repository checks, lint and contracts, all four unit shards, kernel smoke, trajectory checks, and coverage gates also pass.
gloryfromca
left a comment
gloryfromca
left a comment
There was a problem hiding this comment.
No blockers; this can merge as far as I am concerned.
The revision change only merges the latest target branch. I reviewed the delta from 667e52d5fa62, the resulting github/main...HEAD diff, merge history, and the applicable AGENTS.md/CLAUDE.md and CONTEXT-MAP.md constraints. The effective feature diff remains the same one-line README chart swap, with no conflict-resolution change, new caller or architecture impact, compatibility change, or weakened test.
The merge result passes git diff --check and the local large-file, source-language, and commit-message gates. GitHub's full check set is green, including the docs build, all four unit shards, and coverage.
## Summary Adds a second row to the Showcase's non-deck table, bringing the design engine's two output modes into the section: a poster campaign and an interactive page. **Light-pollution poster campaign.** The first case whose graph takes an edge from the code side. Three Raven-Research nodes and a plate-generation node run in parallel; a Raven-Code node computes the sky-brightness panel from the research findings; a Raven-Design node composes the key visual and the street, data-panel, social and banner cuts off one master plate, with every headline and statistic set as live type rather than baked into the generated pixels. The artifact image is the key visual over the four derived formats. **Interactive GPS explainer.** Two Raven-Code nodes write a trilateration solver and an independent oracle in parallel, a Raven-Oncall node cross-checks them, and a Raven-Design node builds the page on a generated orbital plate. The artifact image is the live page above the fix at two, three and four satellites, which is what shows the geometry is computed rather than drawn. Both images are hosted as GitHub user-attachments and referenced by URL. Nothing is committed to the repository, per the repository-assets rule. ## Type - [ ] Fix - [ ] Feature - [x] Docs - [ ] CI / tooling - [ ] Refactor - [ ] Other ## Verification - `make check-large-files` -> exit 0 - `make check-source-language` -> exit 0 - All four attachment URLs fetched anonymously with curl: HTTP 200, and each one's md5 compared against its local source file to confirm the right image sits behind the right URL. - [ ] Relevant tests pass locally - [x] Relevant lint / type checks pass locally - [x] User-facing docs or screenshots are updated when needed ## Risk README-only change. The two new cells add four remote images to the page; if an attachment URL ever stops resolving the cell degrades to alt text. Rollback is reverting the commit. - [ ] Security impact considered - [x] Backward compatibility considered - [x] Rollback path is clear for risky changes ## Related Issues Follows #517 and #520. --------- Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
3 participants
Summary
Swaps the sweep chart in the Showcase for an annotated version.
The chart that shipped with #517 was the run's raw matplotlib output: four
flat lines and four noisy ones, with the finding left for the reader to infer.
The new one states it in its own panel titles, marks the best cell, labels the
recall values on the right edge, and carries a footnote saying the source is
18 rows with duplicate cells averaged. That footnote also explains the two
points that are means of two runs, which the earlier chart showed without
comment.
The image is padded to the same aspect as the comparison board beside it, on
the chart's own white canvas, so the two cells in that table still end at the
same height. Hosted as a GitHub user attachment, like the rest of the section.
Type
Verification
make check-large-filesandmake check-source-languageboth pass; thechange is one line of README markup.
Confirmed the new attachment returns HTTP 200 to an anonymous request and
matches its source file by checksum.
Rendered the section locally at GitHub's 860px README column width to check
that the two cells in the second table still align.
Relevant tests pass locally
Relevant lint / type checks pass locally
User-facing docs or screenshots are updated when needed
No tests or lint were run: this change touches README markup only.
Risk
README-only, with no runtime behaviour change. Rollback is reverting the
commit; the previous image is still a live attachment, so the old chart can
come back by restoring the one line.
Related Issues
Follows #517.