CI: push to ipcr-push-85-17-246-67.ip.nsl.sh (the box's cert no longer covers nip.io) [skip ci]

Talk, news, soap and reality air every day; a row bounded by "aired today" is
otherwise their schedule. Excluded on the airing path only — a popular row has
no reason to hide a talk show.

.gitea/workflows/publish.yml

@@ -0,0 +1,88 @@
+# Gitea twin of .github/workflows/publish.yml. Gitea runs only .gitea/workflows/ once this
+# folder exists, and GitHub never reads it.
+#
+# Publishes to IPCR on watch.nsl.sh instead of ghcr.io:
+#   push  ipcr-push-85-17-246-67.ip.nsl.sh/metadec.eth/<repo>:<tag>   docker login as metadec.eth with the
+#         IPCR_PUSH_TOKEN secret (the metadec.eth namespace token from the IPCR dashboard)
+#   pull  ipcr-watch.nsl.sh/ipns/metadec.eth/metadec.eth/<repo>:<tag>   anonymous
+# The push goes to ipcr-push-85-17-246-67.ip.nsl.sh, the box's own address, not ipcr-push-watch.nsl.sh: Cloudflare fronts
+# *.watch.nsl.sh and refuses any request over 100 MB (413), which a large image layer is.
+# That host's certificate is signed by the box's private Mesh Router CA, given to buildkit from
+# the IPCR_CA_PEM repo variable. docker login would make the job's Docker daemon check the
+# certificate without that CA, so the login is written straight into ~/.docker/config.json.
+# The runner's job image (node:22-bookworm) has no docker CLI, so the static one is
+# installed first; buildx installs itself. No `type=gha` cache: the runner has none.
+
+name: Publish image
+
+# Build and push the feeder image to the IPCR (ipcr-push-watch.nsl.sh):
+#   * every push to main  -> the moving `main` tag
+#   * every `v*` tag       -> semver tags (e.g. v1.0.0 -> 1.0.0, 1.0, 1)
+# The build context is the repo root (the workspace; `meta-feeder-sdk` is a
+# git dependency pinned by tag); the Dockerfile lives under
+# feeder-plugin/card-feeder/.
+on:
+  push:
+    branches: [main]
+    tags: ['v*']
+
+env:
+  REGISTRY: ipcr-push-85-17-246-67.ip.nsl.sh
+  IMAGE_NAME: metadec.eth/${{ github.event.repository.name }}
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Install Docker CLI
+        run: curl -fsSL https://download.docker.com/linux/static/stable/x86_64/docker-28.5.1.tgz | tar -xz -C /usr/local/bin --strip-components=1 docker/docker
+
+      - name: Trust the Mesh Router CA
+        run: printf '%s\n' "$IPCR_CA_PEM" > /tmp/ipcr-ca.pem
+        env:
+          IPCR_CA_PEM: ${{ vars.IPCR_CA_PEM }}
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+        with:
+          version: latest
+          buildkitd-config-inline: |
+            [registry."${{ env.REGISTRY }}"]
+              ca = ["/tmp/ipcr-ca.pem"]
+
+      - name: Log in to IPCR
+        env:
+          IPCR_PUSH_TOKEN: ${{ secrets.IPCR_PUSH_TOKEN }}
+        run: |
+          mkdir -p ~/.docker
+          printf '{"auths":{"%s":{"auth":"%s"}}}\n' "$REGISTRY" "$(printf 'metadec.eth:%s' "$IPCR_PUSH_TOKEN" | base64 -w0)" > ~/.docker/config.json
+
+      - name: Extract image metadata
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+          tags: |
+            type=ref,event=branch
+            type=semver,pattern={{version}}
+            # A fix to release 1.2.3 is tagged v1.2.3.1: four numbers are not semver, so
+            # the pattern above makes no tag for it; this one does (image tag 1.2.3.1).
+            type=match,pattern=v(\d+\.\d+\.\d+\.\d+),group=1
+            type=semver,pattern={{major}}.{{minor}}
+            type=semver,pattern={{major}}
+            type=sha,format=short
+
+      - name: Build and push
+        uses: docker/build-push-action@v6
+        with:
+          context: .
+          file: ./feeder-plugin/card-feeder/Dockerfile
+          push: true
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}

.github/workflows/publish.yml

@@ -43,6 +43,9 @@ jobs:
           tags: |
             type=ref,event=branch
             type=semver,pattern={{version}}
+            # A fix to release 1.2.3 is tagged v1.2.3.1: four numbers are not semver, so
+            # the pattern above makes no tag for it; this one does (image tag 1.2.3.1).
+            type=match,pattern=v(\d+\.\d+\.\d+\.\d+),group=1
             type=semver,pattern={{major}}.{{minor}}
             type=semver,pattern={{major}}
             type=sha,format=short

feeder-plugin/card-feeder/src/consts.rs

@@ -47,6 +47,20 @@ pub(crate) const TMDB_DISCOVERY_WAIT_DEADLINE_SECS: u64 = 5;
 /// asking for 500) to a predictable per-row TMDB cost.
 pub(crate) const DISCOVERY_MAX_PAGES: u32 = 5;
 
+/// TMDB genre ids kept OUT of the date-bounded `airing` rows.
+///
+/// A row bounded by "aired today" is otherwise dominated by what airs EVERY
+/// day: late-night talk (`The Tonight Show`, `Watch What Happens Live`), the
+/// news-desk shows (`The Daily Show`), long-running soaps (`Goede Tijden,
+/// Slechte Tijden`, `Demain nous appartient`) and reality. Measured on the live
+/// row: 6 of the first 6 titles. They are legitimately airing — they are simply
+/// not what "what came out today" is asked to answer, and they crowd out the
+/// drama that airs weekly.
+///
+/// 10764 Reality · 10767 Talk · 10763 News · 10766 Soap.
+/// Drop an id from this list to let that genre back into the row.
+pub(crate) const AIRING_EXCLUDED_GENRES: &str = "10764,10767,10763,10766";
+
 /// TMDB keyword ids excluded from an `anime:true` discovery row, as
 /// `without_keywords`: `hentai`, `softcore`, `erotica`, `pornography`,
 /// `adult animation`.

feeder-plugin/card-feeder/src/discovery.rs

@@ -58,7 +58,9 @@ use meta_feeder_sdk::plugin::GatewayQuery;
 use tracing::debug;
 
 use crate::card::{Card, CardSource};
-use crate::consts::{DISCOVERY_EXCLUDED_KEYWORDS, DISCOVERY_MAX_PAGES, DISCOVERY_MIN_VOTES};
+use crate::consts::{
+    AIRING_EXCLUDED_GENRES, DISCOVERY_EXCLUDED_KEYWORDS, DISCOVERY_MAX_PAGES, DISCOVERY_MIN_VOTES,
+};
 use crate::resolve::{dedup_titles, Resolver};
 use crate::tmdb_client::{title_names, TmdbHit, TmdbKind};
 
@@ -77,6 +79,15 @@ pub(crate) enum DiscoveryMode {
     Popular,
     Trending,
     TopRated,
+    /// What came out **today** — the day-by-day row. TV: shows with an episode
+    /// airing today; film: releases from the last fortnight.
+    ///
+    /// Date-bounded rather than availability-bounded on purpose. "Is a file out
+    /// yet" is an indexer question, and asking it per show per day is exactly
+    /// the load that wedges a Prowlarr; an air date is a fact TMDB already
+    /// holds, costs the same single call as every other row, and makes the row
+    /// roll over by itself at midnight with nothing scheduled.
+    Airing,
 }
 
 impl DiscoveryMode {
@@ -94,6 +105,7 @@ impl DiscoveryMode {
             DiscoveryMode::Popular => "popular",
             DiscoveryMode::Trending => "trending",
             DiscoveryMode::TopRated => "top_rated",
+            DiscoveryMode::Airing => "airing",
         }
     }
 
@@ -103,7 +115,11 @@ impl DiscoveryMode {
     fn discover_sort(self) -> &'static str {
         match self {
             DiscoveryMode::TopRated => "vote_average.desc",
-            DiscoveryMode::Popular | DiscoveryMode::Trending => "popularity.desc",
+            // Airing sorts by popularity too: the date bound is what selects the
+            // row, so the sort only decides who leads it.
+            DiscoveryMode::Popular | DiscoveryMode::Trending | DiscoveryMode::Airing => {
+                "popularity.desc"
+            }
         }
     }
 }
@@ -124,6 +140,10 @@ pub(crate) fn is_discovery_query(query: &GatewayQuery) -> bool {
 /// First truthy mode marker on the query, in precedence order.
 fn discovery_mode(query: &GatewayQuery) -> Option<DiscoveryMode> {
     for mode in [
+        // Airing leads: it is the most specific claim a row can make, and a
+        // caller that sends `airing:true popular:true` means "today's, by
+        // popularity", not "popular, ever".
+        DiscoveryMode::Airing,
         DiscoveryMode::Trending,
         DiscoveryMode::Popular,
         DiscoveryMode::TopRated,
@@ -203,6 +223,70 @@ fn with_genres_ids(wanted: &[String], names: &HashMap<u32, String>) -> Vec<u32>
     ids
 }
 
+/// How far back the film side of [`DiscoveryMode::Airing`] reaches, in days.
+///
+/// A series airs an episode on a given day, so "today" is a real set. Films do
+/// not: a single day's releases are usually empty, and a viewer opening the app
+/// on a Tuesday still means "what's new". A fortnight is the smallest window
+/// that keeps the row populated without turning it into "this year".
+const AIRING_MOVIE_WINDOW_DAYS: i64 = 13;
+
+/// Today, UTC, as `YYYY-MM-DD` — the bound [`DiscoveryMode::Airing`] is built
+/// from. UTC rather than a local zone: the feeder serves peers in every zone,
+/// and TMDB's air dates are themselves zone-less calendar dates.
+fn today_utc_days() -> i64 {
+    std::time::SystemTime::now()
+        .duration_since(std::time::UNIX_EPOCH)
+        .map(|d| (d.as_secs() / 86_400) as i64)
+        .unwrap_or(0)
+}
+
+/// `YYYY-MM-DD` for a count of days since 1970-01-01.
+///
+/// Hand-rolled (Howard Hinnant's civil-from-days) rather than pulling in a date
+/// crate for one format call — and split from [`today_utc_days`] so the whole
+/// thing is testable against known dates instead of "whatever today is".
+fn fmt_days(days: i64) -> String {
+    let z = days + 719_468;
+    let era = if z >= 0 { z } else { z - 146_096 } / 146_097;
+    let doe = (z - era * 146_097) as i64; // [0, 146096]
+    let yoe = (doe - doe / 1460 + doe / 36_524 - doe / 146_096) / 365; // [0, 399]
+    let y = yoe + era * 400;
+    let doy = doe - (365 * yoe + yoe / 4 - yoe / 100); // [0, 365]
+    let mp = (5 * doy + 2) / 153; // [0, 11]
+    let d = doy - (153 * mp + 2) / 5 + 1; // [1, 31]
+    let m = if mp < 10 { mp + 3 } else { mp - 9 }; // [1, 12]
+    let y = if m <= 2 { y + 1 } else { y };
+    format!("{y:04}-{m:02}-{d:02}")
+}
+
+/// The `/discover` date bounds for [`DiscoveryMode::Airing`], given today.
+///
+/// TV bounds `air_date` to the single day, which is what makes the row
+/// "today's episodes" and what makes it roll over on its own. Film bounds
+/// `primary_release_date` to [`AIRING_MOVIE_WINDOW_DAYS`] back — see that
+/// constant for why the two kinds cannot share a window.
+fn airing_bounds(kind: TmdbKind, today: i64) -> String {
+    match kind {
+        TmdbKind::Tv => {
+            let d = fmt_days(today);
+            // ⚠ `without_genres` rides with the date bound, and only here. A row
+            // of "aired today" is otherwise the daily-broadcast schedule — talk,
+            // news, soap, reality — which is exactly what the viewer did not
+            // mean. See [`AIRING_EXCLUDED_GENRES`]. The other modes keep every
+            // genre: a *popular* row has no reason to hide a talk show.
+            format!(
+                "air_date.gte={d}&air_date.lte={d}&without_genres={AIRING_EXCLUDED_GENRES}&"
+            )
+        }
+        TmdbKind::Movie => format!(
+            "primary_release_date.gte={}&primary_release_date.lte={}&",
+            fmt_days(today - AIRING_MOVIE_WINDOW_DAYS),
+            fmt_days(today),
+        ),
+    }
+}
+
 /// Build the `path_and_query` for `TmdbClient::discovery_list` from the mode,
 /// kind, anime flag, requested genre ids, and 1-based `page`.
 fn build_path_and_query(
@@ -244,8 +328,18 @@ fn build_path_and_query(
     // single 10.0 vote on top, which a 50-vote gate doesn't stop.
     let votes = match mode {
         DiscoveryMode::TopRated => 200,
+        // ⚠ No vote floor on the airing row, and that is the point. The shared
+        // gate exists to keep unrated residue off a popularity ranking; here the
+        // date bound already decides membership, and a floor would filter out
+        // exactly what the row is for — a series whose first season is airing
+        // now has not had time to collect 50 votes.
+        DiscoveryMode::Airing => 0,
         _ => DISCOVERY_MIN_VOTES,
     };
+    let dates = match mode {
+        DiscoveryMode::Airing => airing_bounds(kind, today_utc_days()),
+        _ => String::new(),
+    };
     let keyword = if anime {
         format!("with_keywords={TMDB_ANIME_KEYWORD}&")
     } else {
@@ -266,7 +360,7 @@ fn build_path_and_query(
         format!("with_genres={joined}&")
     };
     format!(
-        "discover/{seg}?{keyword}{genres}sort_by={}&include_adult=false\
+        "discover/{seg}?{keyword}{genres}{dates}sort_by={}&include_adult=false\
          &without_keywords={DISCOVERY_EXCLUDED_KEYWORDS}&vote_count.gte={votes}&page={page}",
         mode.discover_sort(),
     )
@@ -719,4 +813,56 @@ mod tests {
         let movie = card_from_hit(&hit(), TmdbKind::Movie, &genre_table());
         assert_eq!(movie.content_kind(), "movie");
     }
+
+    /// The day-by-day row: what selects it, and what it asks TMDB for.
+    ///
+    /// Bounding by air date rather than by availability is the whole design —
+    /// see [`DiscoveryMode::Airing`]. These pin the three parts that make the
+    /// row roll over on its own: the date maths, the bounds per kind, and the
+    /// fact that the vote floor is lifted (a season airing now has no votes
+    /// yet, and the shared floor would empty the row it is meant to fill).
+    #[test]
+    fn airing_row_is_bounded_by_date_not_by_votes() {
+        // Date maths, against known days-since-epoch.
+        assert_eq!(fmt_days(0), "1970-01-01");
+        assert_eq!(fmt_days(19_723), "2024-01-01"); // a leap year's first day
+        assert_eq!(fmt_days(19_782), "2024-02-29"); // the leap day itself
+        assert_eq!(fmt_days(20_453), "2025-12-31");
+
+        // TV: exactly one day, so "today's episodes" is literal.
+        let tv = airing_bounds(TmdbKind::Tv, 19_723);
+        assert!(tv.starts_with("air_date.gte=2024-01-01&air_date.lte=2024-01-01&"), "{tv}");
+        // The daily-broadcast genres are excluded from this row, and only this
+        // row — a talk show airs every day and would otherwise fill it.
+        assert!(tv.contains(&format!("without_genres={AIRING_EXCLUDED_GENRES}")), "{tv}");
+        assert!(
+            !build_path_and_query(DiscoveryMode::Popular, TmdbKind::Tv, false, &[], 1)
+                .contains("without_genres"),
+            "a popular row must keep every genre"
+        );
+
+        // Film: a fortnight back, because a single day of releases is empty.
+        let movie = airing_bounds(TmdbKind::Movie, 19_723);
+        assert!(movie.contains("primary_release_date.gte=2023-12-19"), "{movie}");
+        assert!(movie.contains("primary_release_date.lte=2024-01-01"), "{movie}");
+
+        // The marker selects the mode, and outranks a co-sent `popular`.
+        assert_eq!(
+            discovery_mode(&query(&[("airing", "true"), ("popular", "true")])),
+            Some(DiscoveryMode::Airing)
+        );
+
+        // The built query carries the date bound and lifts the vote floor.
+        let url = build_path_and_query(DiscoveryMode::Airing, TmdbKind::Tv, false, &[], 1);
+        assert!(url.starts_with("discover/tv?"), "{url}");
+        assert!(url.contains("air_date.gte="), "{url}");
+        assert!(url.contains("vote_count.gte=0"), "{url}");
+        // Still gated for adult residue like every other /discover row.
+        assert!(url.contains("without_keywords="), "{url}");
+
+        // A popular row is unchanged: no date bound, floor intact.
+        let pop = build_path_and_query(DiscoveryMode::Popular, TmdbKind::Tv, false, &[], 1);
+        assert!(!pop.contains("air_date."), "{pop}");
+        assert!(pop.contains(&format!("vote_count.gte={DISCOVERY_MIN_VOTES}")), "{pop}");
+    }
 }