Phase 4: the site the archive deserves

.github/workflows/gh-pages.yml

@@ -45,6 +45,14 @@ jobs:
       - name: Build
         run: hugo --minify
 
+      - name: Build the search index
+        # Pagefind reads the pages Hugo wrote and writes public/pagefind/:
+        # the engine, the word index in chunks, one fragment per page. It
+        # runs after Hugo because it indexes what is on disk, and it is
+        # pinned the way Hugo is: bin/install-pagefind fetches the release
+        # named in .pagefind-version and checks its SHA-256.
+        run: bin/install-pagefind && bin/search public
+
       - name: Deploy
         uses: peaceiris/actions-gh-pages@84c30a85c19949d7eee79c4ff27748b70285e453 # v4.1.0
         with:

.github/workflows/links.yml

@@ -0,0 +1,60 @@
+# Check the external links of the site.
+#
+# Not part of the pull request checks. A link to another site breaks when
+# that site changes, which is not something a pull request of ours causes
+# or can fix, and asking for 139 external URLs on every pull request is
+# both slow and close to behaving like a crawler. Run this when you want
+# to know, which in practice is before a round of archive work.
+#
+# It never turns anything red. Read the report in the job summary.
+name: links
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  external:
+    name: Check the external links
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+
+      - name: Read the Hugo version
+        run: echo "HUGO_VERSION=$(tr -d '[:space:]' < .hugo-version)" >> "$GITHUB_ENV"
+
+      - name: Setup Hugo
+        uses: peaceiris/actions-hugo@2752ce1d29631191ea3f27c23495fa06139a5b78 # v3.2.1
+        with:
+          hugo-version: ${{ env.HUGO_VERSION }}
+          extended: true
+
+      - name: Build
+        run: hugo --minify -d public --baseURL /
+
+      - name: Check links
+        uses: lycheeverse/lychee-action@e7477775783ea5526144ba13e8db5eec57747ce8 # v2.9.0
+        with:
+          # --exclude-path: the frozen 2016-2019 snapshots under archives/
+          #   keep their dead links on purpose.
+          # --exclude: hosts that answer bots with 400, 403, or 999, and
+          #   canary.invalid, the host the canary fixture names as text.
+          #   lychee reads URLs out of text nodes too, and .invalid never
+          #   resolves.
+          args: >-
+            --no-progress
+            --exclude-all-private
+            --max-concurrency 4
+            --cache
+            --max-cache-age 1w
+            --root-dir "$GITHUB_WORKSPACE/public"
+            --exclude-path 'public/archives/'
+            --exclude 'twitter\.com'
+            --exclude '^https?://(www\.)?x\.com/'
+            --exclude 'linkedin\.com'
+            --exclude 'pybcn\.slack\.com'
+            --exclude 'canary\.invalid'
+            'public/**/*.html'
+          fail: false

.github/workflows/pr.yml

@@ -42,11 +42,21 @@ jobs:
       - name: Build
         run: hugo --minify
 
+      - name: Build the search index
+        # Pagefind reads the pages Hugo wrote and writes public/pagefind/:
+        # the engine, the word index in chunks, one fragment per page. It
+        # runs after Hugo because it indexes what is on disk, and it is
+        # pinned the way Hugo is: bin/install-pagefind fetches the release
+        # named in .pagefind-version and checks its SHA-256.
+        run: bin/install-pagefind && bin/search public
+
   content:
     name: Check the content
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+        with:
+          fetch-depth: 0    # bin/authors --check reads the whole history
 
       - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
         with:
@@ -64,6 +74,19 @@ jobs:
       - name: Check content for dangerous raw HTML
         run: bin/check-html-safety
 
+      - name: Check that AUTHORS is what bin/authors writes
+        # The check writes the file again in memory and compares, so a new
+        # contributor, a second account of somebody already here, and a name
+        # that changed on a person record are all caught. It reads the
+        # GitHub API for the account of each commit and for the reviews,
+        # with the token the job already has; when the API cannot be reached
+        # it says it skipped and passes, because an outage of somebody else
+        # is not a broken build. The reviews come through GraphQL, three
+        # requests for the 205 pull requests, which is why it takes seconds.
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: bin/authors --check
+
       - name: Check the text normalizer
         run: python3 bin/canon.py --selftest
 
@@ -117,29 +140,28 @@ jobs:
         # runner's python3 is enough.
         run: bin/check-rendered public
 
-      - name: Check links
+      - name: Check the internal links
         uses: lycheeverse/lychee-action@e7477775783ea5526144ba13e8db5eec57747ce8 # v2.9.0
         with:
-          # --root-dir: where "/path/" links resolve. Without it lychee flags
-          #   every absolute link in a local file as an error.
+          # --offline: only the links that point inside this build. The
+          #   external ones are the job of .github/workflows/links.yml,
+          #   which runs on request. The site holds 139 distinct external
+          #   URLs outside the frozen snapshots, and checking them took 25
+          #   minutes on the pull request that added the person pages: with
+          #   the concurrency low enough not to look like a crawler, every
+          #   pull request asked every host we have ever linked to whether
+          #   it is still there, and reported rot that is not ours to fix.
+          #   An internal link that does not resolve is a typo, is ours,
+          #   and is found in seconds.
+          # --root-dir: where "/path/" links resolve. Without it lychee
+          #   flags every absolute link in a local file as an error.
           # --exclude-path: the frozen 2016-2019 snapshots under archives/
           #   keep their dead links on purpose.
-          # --exclude: hosts that answer bots with 400, 403, or 999, and
-          #   canary.invalid, the host the canary fixture names as text.
-          #   lychee reads URLs out of text nodes too, and .invalid never
-          #   resolves.
           args: >-
             --no-progress
-            --exclude-all-private
-            --max-concurrency 4
+            --offline
             --root-dir "$GITHUB_WORKSPACE/public"
             --exclude-path 'public/archives/'
-            --exclude 'twitter\.com'
-            --exclude '^https?://(www\.)?x\.com/'
-            --exclude 'linkedin\.com'
-            --exclude 'pybcn\.slack\.com'
-            --exclude 'canary\.invalid'
             'public/**/*.html'
-          # External sites rate-limit. Keep the report in the job summary and
-          # do not turn the check red.
-          fail: false
+          # An internal link is ours, so this one does turn the check red.
+          fail: true

.gitignore

@@ -1,5 +1,6 @@
 .idea
 bin/hugo
+bin/pagefind
 public/**
 
 # Hugo build output and caches. resources/ covers the SCSS cache and the

.pagefind-version

@@ -0,0 +1 @@
+1.5.2

AGENTS.md

@@ -14,22 +14,35 @@ bin/install     # downloads the pinned Hugo binary, verifies its SHA256
 bin/serve       # dev server, Hugo's default :1313
 bin/build       # writes to docs/
 bin/hugo --quiet -d /tmp/out     # the binary directly
+bin/install-pagefind             # the pinned Pagefind binary, verifies its SHA256
+bin/search /tmp/out              # the search index of a built site, after the build
 ```
 
 **The Hugo binary lives at `bin/hugo` and is gitignored.** `bin/helpers` prefers
 it over anything on `$PATH`, and checks the version from `.hugo-version`.
 
+**The search index is not built by Hugo.** `bin/search <built-dir>` runs
+Pagefind (pinned in `.pagefind-version`, binary at the gitignored
+`bin/pagefind`) over the pages Hugo wrote, and CI runs it right after
+`hugo --minify`. The dev server has none, so `/search/` says so there. What
+is indexed is what the templates mark `data-pagefind-body`: the record of
+each talk, event and person, and the photo captions, not the whole page.
+`partials/search_kind.html` says which pages, and `partials/hero_header.html`
+and the layouts of the three kinds say what of each.
+
 **Do not `pip install hugo`.** That PyPI package now requires a Go toolchain and
 fails. `bin/install` fetches the official release binary instead.
 
-A clean build with `-D` produces **1369 HTML pages**: 1313 Hugo-generated and
+A clean build with `-D` produces **1377 HTML pages**: 1321 Hugo-generated and
 56 copied from `static/archives/`. Recount rather than trusting this line: the
 number moves with the content. It was 317 until #202 stopped publishing the
 206 person and sponsor pages that nothing linked to, 109 until the archive
-import added a page per meetup and per talk, and 991 until the person pages
+import added a page per meetup and per talk, 991 until the person pages
 came back with the Archive (378 of them, one per record of `content/people/`,
-the canary fixture included), which is how quickly a figure like this goes
-stale. Recount with:
+the canary fixture included), and 1363 until the tabs of `/talks/` and
+`/people/` added a page per talk type and per role (the two content adapters
+under `content/`), which is how quickly a figure like this goes stale.
+Recount with:
 
 ```sh
 find <built-dir> -name '*.html' | wc -l
@@ -45,6 +58,26 @@ and checking the output, not by reading the file.
 **`merge.ff = only`.** Any local merge needs `--no-ff`. A failed merge here reports
 "Not possible to fast-forward", which is not a conflict.
 
+**A photograph belongs to the event its text names, not to the day it was
+posted.** `bin/import-x-photos --ids FILE` takes the posts to import, and the
+day stops deciding. Reading the export by date looks right and is wrong three
+ways, each found the hard way on 2026-10-17:
+
+- The account replies to its own posts of years before, so following the chain
+  of replies builds threads that run from 2014 to 2021. Cut the chain at four
+  days from the photograph.
+- A talk titled in two or three generic words matches any post that uses them:
+  a 2014 talk swallowed every post that said "machine learning" since. A title
+  has to be long and belong to one event before it can attribute on its own.
+- Some titles repeat anyway. There are several "Introduction to Python", and
+  "Barcelona Python Meetup" is both the name of the event of 2007 and what
+  every monthly meetup is called. There the text says what the event is and
+  the day says which one of them, which is the rule a person would use.
+
+A post with photographs and no text at all is not imported: the caption is the
+alt text of the image, and a gallery leaves a photograph out rather than
+publish one nobody can read.
+
 **The archive imports run in one order and only once.** `bin/import-meetups`,
 then `bin/import-talks`, then `bin/import-videos`. `bin/import-talks` replaces
 the whole `talks` block of a page and everything after it in the front matter,
@@ -219,8 +252,9 @@ plausible but different result.
 | `themes/pybcn_theme/layouts/partials/` | Most of the interesting template logic |
 | `static/archives/` | Verbatim mirrors of pre-Hugo sites, copied unprocessed |
 | `data/` | Tables the templates read: `roles.yaml` maps a heading to a role, `role_emoji.yaml` a role to a glyph, `event_emoji.yaml` an event to a glyph, `venues.yaml` the exact `venue` string of an event to the short name the `/events/` list prints (the page of the meetup keeps the full string). Plus `channel_videos.yaml`, the inventory of the YouTube channel, which `bin/check-content` reads and `bin/list-channel` refreshes |
-| `bin/` | Setup and build scripts, the checks, `bin/list-channel` (names a channel video the inventory lacks; needs `yt-dlp` and the network), the one-time `bin/import-*` scripts, plus the gitignored Hugo binary |
+| `bin/` | Setup and build scripts, the checks, `bin/list-channel` (names a channel video the inventory lacks; needs `yt-dlp` and the network), `bin/new-meetup` (writes the page of one past meetup from its Meetup URL, and `--discover` names the ones with no page; needs the network), the one-time `bin/import-*` scripts, plus the gitignored Hugo binary |
 | `.hugo-version` | The pinned Hugo version, single source of truth |
+| `.pagefind-version` | The pinned Pagefind version, same rule; `bin/install-pagefind` reads it |
 
 The checks, which are what a pull request has to pass:
 

AUTHORS

@@ -0,0 +1,52 @@
+PyBCN publishes this site. These are the people who built it, in the order
+they arrived. Writing it and reviewing it both count, because a change that
+nobody reads is not finished, and the years are the first and the last one
+each person worked on it.
+
+The name of each one is the name their page of the site carries, or the name
+they signed their own commits with when they have no page. A person who would
+rather appear another way can change it in a pull request, because this file
+is generated: from the commits, from the reviews, and from the pages.
+
+The people behind the events themselves, the speakers, the organizers, the
+venues and the sponsors, are named on the pages that stay current, which
+static/humans.txt links.
+
+Written by bin/authors.
+
+Areski Belaid (@areski) wrote, 2013 to 2015.
+    https://pybcn.org/people/areski-belaid/
+Núria Pujol Vilanova (@llevaNEUS) wrote, 2014 to 2015.
+    https://pybcn.org/people/nuria-pujol-vilanova/
+David Arcos (@DZPM) wrote and reviewed, 2015 to 2026.
+    https://pybcn.org/people/david-arcos/
+Jordi Soucheiron (@jsoucheiron) wrote, 2015 to 2016.
+    https://pybcn.org/people/jordi-soucheiron/
+Loris Mularoni (@batterio) wrote, 2015.
+    https://pybcn.org/people/loris-mularoni/
+Ignasi Fosch (@ifosch) wrote and reviewed, 2016 to 2026.
+    https://pybcn.org/people/ignasi-fosch/
+Laura Pérez Mayos (@lpmayos) wrote and reviewed, 2016 to 2025.
+    https://pybcn.org/people/laura-perez-mayos/
+Israel Saeta Pérez (@dukebody) wrote and reviewed, 2018 to 2024.
+    https://pybcn.org/people/israel-saeta-perez/
+Ruben Berenguel (@rberenguel) wrote and reviewed, 2018 to 2023.
+    https://pybcn.org/people/ruben-berenguel/
+Juan Luis Cano Rodríguez (@astrojuanlu) reviewed, 2018.
+    https://pybcn.org/people/juan-luis-cano-rodriguez/
+Natàlia Padilla (@NataliaSirera) wrote and reviewed, 2019 to 2025.
+    https://pybcn.org/people/natalia-padilla/
+Dimitri Herrero (@dim1tri) wrote, 2019.
+    https://pybcn.org/people/dimitri-herrero/
+Xavi Francisco Gilabert (@srxavi) wrote and reviewed, 2021 to 2022.
+    https://pybcn.org/people/xavi-francisco-gilabert/
+Ferran Jovell (@mrswats) wrote and reviewed, 2023 to 2024.
+    https://pybcn.org/people/ferran-jovell/
+Alicia Morales Carrasco (@amoralca16) wrote and reviewed, 2024 to 2025.
+    https://pybcn.org/people/alicia-morales-carrasco/
+Sergi Ramírez (@ramitjansIA, @ramitjans) wrote, 2025.
+    https://pybcn.org/people/sergi-ramirez/
+Daniel Mesejo León (@mesejo) wrote and reviewed, 2025 to 2026.
+    https://pybcn.org/people/daniel-mesejo-leon/
+Alberto Cámara (@ber2) reviewed, 2026.
+    https://pybcn.org/people/alberto-camara/

CONTRIBUTING.md

@@ -97,8 +97,18 @@ Edit the file on GitHub, or clone the repository and follow the steps below.
   [Person photos](README.md#person-photos).
 - **A sponsor**: run `bin/hugo new sponsors/my-sponsor.md` and fill the four
   fields listed in [How to add a new sponsor](README.md#how-to-add-a-new-sponsor).
-  The logo goes under `themes/pybcn_theme/assets/images/sponsors/`. Then list
-  the sponsor on the sponsors page, or on the event, that shows it.
+  The logo goes under `themes/pybcn_theme/assets/images/sponsors/`. The
+  [sponsors page](https://pybcn.org/sponsors/) reads the directory, so the
+  new file reaches it by existing; list the sponsor on its event as well, in
+  the `sponsor_levels` of that page, which is what gives it a level.
+- **A monthly meetup**, after it happened: run `bin/new-meetup --discover`
+  to see which ones have no page yet, then
+  `bin/new-meetup <the Meetup URL of the event>`. It writes the page from the
+  public event page: the title, the day, the venue, the number of
+  registrations, the source URL, and the group. It writes no talk, because the
+  announcement says them in prose: pass `--description` to print that prose,
+  and see [Import a meetup](README.md#import-a-meetup) for the rest of the
+  path, which is the same one the 268 pages of the archive came through.
 - **An event**: copy the previous edition under `content/events/`, for example
   `content/events/pyday_bcn/pyday_bcn_2025.md`, and edit it. The sections, the
   agenda and the people grids are described in

LICENSING.md

@@ -157,7 +157,7 @@ case (section 2(b)(2) of the legal code).
 
 The PyBCN logo and its variants (`themes/pybcn_theme/assets/images/logo.png`,
 `themes/pybcn_theme/assets/images/favicon.png`, `static/favicon.ico`,
-`static/favicon.png`, and the files under `static/images/backgrounds/`)
+and the files under `static/images/backgrounds/`)
 identify the association. They are not covered by the content licence: use
 them to refer to PyBCN, not to suggest that PyBCN endorses something it has
 not seen. The PyLadies name and logo in that same folder are not the
@@ -174,8 +174,8 @@ with it:
 - `themes/pybcn_theme/static/vendor/font-awesome/webfonts/`: the Font Awesome
   fonts (SIL OFL 1.1).
 - `themes/pybcn_theme/assets/css/bootstrap.min.css`: Bootstrap 4.0.0 (MIT).
-- `themes/pybcn_theme/assets/images/anon_member.png`: the Font Awesome `user`
-  icon as a PNG (CC BY 4.0).
+- `themes/pybcn_theme/assets/images/anon_member.webp`: the Font Awesome `user`
+  icon as a WebP (CC BY 4.0).
 
 ### Archived sites