Repository navigation
Docs: readable evidence, rendered math with scope, remove audit page, add downloads badge - #19
Conversation
Matches the badge the other finite-sample packages carry. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011pXaJ5DLex7BrdUyfTqjL4
It read as an internal changelog of past defects rather than user documentation. The only link to it was the docs toctree. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011pXaJ5DLex7BrdUyfTqjL4
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Enable MyST dollarmath/amsmath so equations render with MathJax instead of plain-text blocks, and rewrite the page in LaTeX. Formalize the per-update decay mode, which the page skipped, and add a Scope section separating what is proved exactly, what it depends on, and what is not claimed (unbinned equivalence, statistical guarantees, drift tracking, outcome-dependent label arrival). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011pXaJ5DLex7BrdUyfTqjL4
Render Brier differences as percentage changes with a plain verdict, known-truth errors in percentage points, sizes in B/KB and latencies rounded to what matters, with method names instead of config strings. Show saved size in the main table, since River's dict state reads as 0 bytes of arrays. Fix the resource table's extra alignment column. Rewrite the prose to lead with findings. Tables are regenerated from the same committed JSON; no benchmark was rerun. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011pXaJ5DLex7BrdUyfTqjL4
soodoku
changed the title
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e10cb1a146
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| if point["test_resource_feasible"]: | ||
| held = "yes" | ||
| else: | ||
| held = f"no ({point['test']['p95_update_ms']:.2f} ms)" |
There was a problem hiding this comment.
When the validation budget admits no method, electricity() in benchmarks/suite.py deliberately records selected, test, and test_resource_feasible as None. This falsey branch then indexes point["test"], causing python -m benchmarks.render to fail with a TypeError instead of emitting the intended “nothing fits” row. Handle the None case before formatting the failed-budget detail.
Useful? React with 👍 / 👎.
Summary
benchmarks/render.py,docs/evidence.md):dollarmath/amsmath, so equations render with MathJax instead of plain-text blocks.docs/audit.md(/audit.html) and its toctree entry.Validation
sphinx-build -Wpasses. In headless Chromium, 77 math elements render with no raw LaTeX left. The built site has noaudit.htmland no links to it.ruff checkandruff format --checkpass;pytest: 144 passed. Both badge URLs return 200.🤖 Generated with Claude Code
https://claude.ai/code/session_011pXaJ5DLex7BrdUyfTqjL4