Release 0.77.12 - #3835
Merged
Merged
Release 0.77.12#3835
Conversation
* Add the content-engagement panel to the B2B analytics dashboard The fifth org-scoped endpoint had a query factory, types and a factory but no UI, so `mv_b2b_content_engagement_depth` was reaching the browser and being dropped on the floor. It is a table rather than a chart. The view mixes counts of people with counts of events that run orders of magnitude larger, which is the same reason those event totals were kept off the trend chart — plotting them together needs a dual axis, and a dual axis invites a comparison the scales do not support. Eleven metrics will not fit as eleven columns at this table's density, so each activity column leads with the per-engaged-learner rate (comparable across course runs of different sizes) and prints the raw total under it. Both halves are suppressed independently by the k-anonymity floor, so both go through SuppressibleValue. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VpA41z1nkQJ9FpbJVd82h1 * Say "per engaged learner" where the average is over engaged learners Read mv_b2b_content_engagement_depth.sql rather than inferring the denominators from the test factory. The two rate families do not share a base: engagement_rate_pct and chatbot_adoption_pct divide by total_enrolled_learners, but the two averages divide by engaged_learners, matching their field names. "Videos per learner" therefore reads as per enrolled learner and overstates the figure on any run where engagement is weak. The "of enrolled" copy on the engaged cell was already correct and stands. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VpA41z1nkQJ9FpbJVd82h1 * Show the cohort behind each activity total ol-analytics-api #33 returns video_watchers and problem_attempters, the distinct learners behind total_videos_watched and total_problems_attempted. The type mirrors that model column for column, so it gains both. They are worth rendering, not just carrying. A bare activity total invites the reader to divide it by the learners they can see, and that is the wrong denominator: only the learners who did that particular thing contributed, which is narrower than engaged_learners. It is also the cohort the anonymity floor is applied to, so showing it is what makes a suppressed total legible rather than arbitrary. The chatbot cell already read this way; videos and problems now match it. Below md the label moves above its value. The shared cell layout sets them side by side, which a label as long as "Problems per engaged learner" plus "96 learners, 9,134 attempted" broke across four lines at 390px. The suppression test now models how the API actually withholds — a sub-floor cohort takes its total and its average with it, while a healthier activity in the same row keeps all three figures — rather than an arbitrary mix of nulls. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VpA41z1nkQJ9FpbJVd82h1 --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
The app shipped no resource hints, so three cross-origin hosts were discovered far too late to open their own connections: - api.learn.mit.edu: the first API call is not issued until React hydrates (~1.2s into a production load), and then waits ~180ms for DNS + TCP + TLS while every sibling API request stalls behind it. - p.typekit.net: not named anywhere in the document. It is only discovered once use.typekit.net's stylesheet downloads and parses. - use.typekit.net: the stylesheet's connection is not reused for font fetches, so the font files paid a fresh TCP + TLS of their own. Each hint's crossOrigin has to match how that resource is actually fetched, because socket pools do not mix across CORS and credentials modes: absent for p.typekit.net's non-CORS stylesheet, anonymous for the CORS-without-credentials font files, and use-credentials for the API, which axios calls with withCredentials. A mismatch is silent -- the socket opens, nothing draws from it, and the request pays the full handshake anyway. Verified in Chrome 151 by reading netlog socket groups rather than request timing, since a preconnect is not reliably observable from timing (RTT variance swamps the effect). Against a same-site origin the API request reuses the preconnected socket -- one connect job, no second handshake -- and its pre-send stall drops from ~180ms to ~1ms. Two cases are partial, both inherent rather than fixable here. React hoists stylesheets above other head links, so use.typekit.net's stylesheet request beats its own hint and only the font files benefit. And for users who block third-party cookies, p.typekit.net's stylesheet lands in a partitioned socket pool that no crossOrigin value targets, leaving the DNS lookup as the only saving there. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
OpenAPI Changes21 changes: 0 error, 0 warning, 21 info Unexpected changes? Ensure your branch is up-to-date with |
odlbot
added
waiting for checkboxes
all checkboxes checked
and removed
deploying to rc
waiting for checkboxes
labels
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
4 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.
zawan-ila
Chris Chudzicki
Tobias Macey