Merge feature/tui-ux into main

TUI navigation and readability improvements on top of the TPB
supplement sources.

README.md

@@ -26,10 +26,12 @@ Sonarr + Radarr + Prowlarr is three separate services, three databases, three we
 
 - **nyaa.si** (RSS) — anime TV, polled continuously. Full-auto, no request budget concerns (it's a plain RSS feed).
 - **apibay.org** (a community JSON API mirror of The Pirate Bay's search) — the primary search-driven source for general TV and movies. Unlike 1337x this is a genuine machine-readable API, needs no HTML scraping, and its search actually ranks by relevance rather than pure seeder count, which matters a lot for titles made of common words.
-- **1337x** (scraped HTML, via community mirrors) — secondary search-driven source, tried after TPB. 1337x's main domain is Cloudflare-protected and has a ban history, so requests are round-robined across mirrors with automatic cooldown/backoff on failures, jittered between searches, and rate-limit responses are honored explicitly.
+- **torrents-csv** (JSON DHT-dump search) — first general-content fallback when TPB fails or returns nothing. Same hash-to-magnet grab as TPB; covers movies and TV.
-- **nyaa.si search mode** — anime movies specifically route here instead of TPB/1337x, since nyaa is the safe, official-RSS-interface target and gives materially better results for anime content.
+- **YTS** (JSON `list_movies` API via `yts.lt` — `yts.mx` no longer resolves) — movie-only fallback after TPB. Each hit expands into one candidate per quality so the scorer sees 720p/1080p/2160p separately.
+- **1337x** (scraped HTML, via community mirrors) — last-resort general-content fallback after the JSON sources. 1337x's main domain is Cloudflare-protected and has a ban history, so requests are round-robined across mirrors with automatic cooldown/backoff on failures, jittered between searches, and rate-limit responses are honored explicitly.
+- **nyaa.si search mode** — anime movies specifically route here instead of the general-content chain, since nyaa is the safe, official-RSS-interface target and gives materially better results for anime content.
 
-All three search-driven sources share one per-cycle request budget (default: 5 searches per 30-minute cycle, TPB tried first, then 1337x, then nyaa search) — kept conservative since 1337x has a ban history and the goal is steady backlog clearing, not maximum throughput. A whole cycle failing outright backs off the *next cycle's* interval (1h → 2h → 4h, capped), on top of each source's own per-mirror cooldowns.
+Search-driven sources share one per-cycle request budget (default: 5 searches per 30-minute cycle). General content tries TPB, then YTS (movies) / torrents-csv, then 1337x; anime movies go to nyaa search. Kept conservative since 1337x has a ban history and the goal is steady backlog clearing, not maximum throughput. A whole cycle failing outright backs off the *next cycle's* interval (1h → 2h → 4h, capped), on top of each source's own per-mirror cooldowns.
 
 ## Setup
 
@@ -52,13 +54,13 @@ Install via bakery (`bakery install breadarr`) on a homelab host, or build from
 
 ## Using the TUI
 
-`Tab` cycles Library / History / Review Queue / Add Show / Stuck / Calendar / Health / Profiles. `j`/`k` or arrow keys navigate, `Enter` opens detail or runs a search, `Esc` backs out.
+`Tab` / `Shift+Tab` cycle Library / History / Review / Add / Stuck / Calendar / Health / Profiles; `1`–`8` jump straight to a tab. `j`/`k` or arrows move, `g`/`G` jump to first/last, `PgUp`/`PgDn` page, `Enter` opens detail or runs a search, `Esc` backs out. `?` is the full key list.
 
-- **Review Queue** — `a` approves and `r` rejects a pending low-confidence title match. Check this periodically, especially early on.
+- **Review** — `a` approves and `r` rejects a pending low-confidence title match. The tab title badges the pending count. Check this periodically, especially early on.
-- **Library** (with an item's detail open) — `s` triggers an immediate search-now pass for that item's backlog; `m`/`e`/`S` toggle monitored on the show/episode/season respectively; `x` (confirm with a second `x`) removes the item from tracking without touching files on disk; `d` (confirm with a second `d`) deletes a bad imported file from disk and clears its tracking, freeing the episode/movie to be re-grabbed on the next cycle — the redownload path for a file that turned out to be wrong or broken; `c` fetches the manual release picker for the selected episode/movie, `Enter` grabs the highlighted candidate, `Esc` cancels.
+- **Library** — `/` incrementally filters the list by title; `f`/`o` cycle kind-filter and sort. With an item's detail open: `s` triggers an immediate search-now pass for that item's backlog; `n` jumps to the next missing monitored episode; `m`/`e`/`S` toggle monitored on the show/episode/season respectively; `x` (confirm with a second `x`) removes the item from tracking without touching files on disk; `d` (confirm with a second `d`) deletes a bad imported file from disk and clears its tracking, freeing the episode/movie to be re-grabbed on the next cycle — the redownload path for a file that turned out to be wrong or broken; `c` fetches the manual release picker for the selected episode/movie, `Enter` grabs the highlighted candidate, `Esc` cancels. Movies (no episode list) show a summary pane with the same keys.
-- **Stuck** — surfaces grabs that look stalled (no download progress advancing, or missing from qBittorrent) before the daemon's own auto-fail timers would catch them.
+- **Stuck** — surfaces grabs that look stalled (no download progress advancing, or missing from qBittorrent) before the daemon's own auto-fail timers would catch them. `Enter` jumps to that show.
-- **Calendar** — upcoming/recently-aired episodes in a roughly week-either-side window.
+- **Calendar** — upcoming/recently-aired episodes in a roughly week-either-side window. `Enter` opens the matching episode.
-- **Health** — the library-health report (see below) rendered as a tab instead of curled by hand.
+- **Health** — daemon cycle outcomes plus the library-health report (see below), scrollable.
 - **Profiles** — quality-profile weight axes; open a profile and edit a weight in place.
 
 ## Operational notes

breadarr-shared/src/config.rs

@@ -97,6 +97,16 @@ pub struct SourcesConfig {
     /// titles made of common words.
     #[serde(default = "default_tpb_api_url")]
     pub tpb_api_url: String,
+    /// torrents.csv DHT-dump search — first fallback when TPB fails or
+    /// returns nothing. Same hash-to-magnet grab shape as TPB, covers
+    /// movies and TV.
+    #[serde(default = "default_torrents_csv_url")]
+    pub torrents_csv_url: String,
+    /// YTS v2 list_movies JSON — movie-only fallback after TPB / torrents-csv.
+    /// `yts.mx` does not resolve; this default is a working host as of
+    /// 2026-08-16.
+    #[serde(default = "default_yts_api_url")]
+    pub yts_api_url: String,
     /// Human kill switch for the upgrade-search loop, same reasoning as
     /// `search_enabled` — off by default would mean nothing ever improves,
     /// but a user who's happy with their current files (or wants to save
@@ -134,6 +144,8 @@ impl Default for SourcesConfig {
             search_budget_per_cycle: default_search_budget_per_cycle(),
             search_enabled: default_search_enabled(),
             tpb_api_url: default_tpb_api_url(),
+            torrents_csv_url: default_torrents_csv_url(),
+            yts_api_url: default_yts_api_url(),
             upgrade_enabled: default_upgrade_enabled(),
             upgrade_poll_interval_secs: default_upgrade_poll_interval_secs(),
             upgrade_budget_per_cycle: default_upgrade_budget_per_cycle(),
@@ -146,6 +158,14 @@ fn default_tpb_api_url() -> String {
     "https://apibay.org/q.php".to_string()
 }
 
+fn default_torrents_csv_url() -> String {
+    "https://torrents-csv.com/service/search".to_string()
+}
+
+fn default_yts_api_url() -> String {
+    "https://yts.lt/api/v2/list_movies.json".to_string()
+}
+
 fn default_upgrade_enabled() -> bool {
     true
 }
@@ -192,6 +212,7 @@ fn default_import_poll_interval_secs() -> u64 {
 
 fn default_1337x_mirrors() -> Vec<String> {
     [
+        "https://www.1337xx.to",
         "https://13377x.info",
         "https://13377x.email",
         "https://1337xto.info",

breadarr-tui/src/main.rs

@@ -6,7 +6,7 @@ use std::time::Duration;
 
 use anyhow::Result;
 use breadarr_shared::{Config, DaemonClient};
-use crossterm::event::{self, Event, KeyCode, KeyEventKind};
+use crossterm::event::{self, Event, KeyCode, KeyEvent, KeyEventKind};
 use crossterm::execute;
 use crossterm::terminal::{
     disable_raw_mode, enable_raw_mode, EnterAlternateScreen, LeaveAlternateScreen,
@@ -51,6 +51,7 @@ async fn run(
 
     loop {
         app.poll_background().await;
+        app.expire_status();
 
         if last_refresh.elapsed() >= Duration::from_secs(3) || app.force_refresh {
             app.refresh_active_tab().await;
@@ -65,7 +66,7 @@ async fn run(
                 if key.kind != KeyEventKind::Press {
                     continue;
                 }
-                handle_key(app, key.code, roots).await;
+                handle_key(app, key, roots).await;
                 if app.should_quit {
                     return Ok(());
                 }
@@ -74,7 +75,9 @@ async fn run(
     }
 }
 
-async fn handle_key(app: &mut App, code: KeyCode, roots: &LibraryRoots) {
+async fn handle_key(app: &mut App, key: KeyEvent, roots: &LibraryRoots) {
+    let code = key.code;
+
     // Typing into the add-show search box takes priority over global keys.
     if matches!(app.tab, Tab::Add) && matches!(app.focus, Focus::AddSearchInput) {
         match code {
@@ -86,7 +89,28 @@ async fn handle_key(app: &mut App, code: KeyCode, roots: &LibraryRoots) {
             KeyCode::Esc => {
                 app.add_query.clear();
             }
-            KeyCode::Tab => cycle_tab(app),
+            KeyCode::Tab => switch_tab(app, tab_offset(app.tab, 1)),
+            KeyCode::BackTab => switch_tab(app, tab_offset(app.tab, -1)),
+            _ => {}
+        }
+        return;
+    }
+
+    // Incremental Library title filter — same input-mode isolation as Add.
+    if matches!(app.tab, Tab::Library) && matches!(app.focus, Focus::LibraryFilterInput) {
+        match code {
+            KeyCode::Enter => app.confirm_library_filter(),
+            KeyCode::Char(c) => {
+                app.library_query.push(c);
+                app.recompute_library_view();
+            }
+            KeyCode::Backspace => {
+                app.library_query.pop();
+                app.recompute_library_view();
+            }
+            KeyCode::Esc => app.clear_library_filter(),
+            KeyCode::Tab => switch_tab(app, tab_offset(app.tab, 1)),
+            KeyCode::BackTab => switch_tab(app, tab_offset(app.tab, -1)),
             _ => {}
         }
         return;
@@ -131,12 +155,25 @@ async fn handle_key(app: &mut App, code: KeyCode, roots: &LibraryRoots) {
 
     match code {
         KeyCode::Char('q') => app.should_quit = true,
-        KeyCode::Tab => cycle_tab(app),
+        KeyCode::Tab => switch_tab(app, tab_offset(app.tab, 1)),
+        KeyCode::BackTab => switch_tab(app, tab_offset(app.tab, -1)),
+        KeyCode::Char(c) if c.is_ascii_digit() => {
+            if let Some(n) = c.to_digit(10) {
+                if (1..=Tab::ALL.len() as u32).contains(&n) {
+                    switch_tab(app, Tab::ALL[(n as usize) - 1]);
+                }
+            }
+        }
         KeyCode::Char('j') | KeyCode::Down => app.move_selection(1),
         KeyCode::Char('k') | KeyCode::Up => app.move_selection(-1),
+        KeyCode::Char('g') => app.select_edge(false),
+        KeyCode::Char('G') => app.select_edge(true),
+        KeyCode::PageDown => app.move_selection(10),
+        KeyCode::PageUp => app.move_selection(-10),
         KeyCode::Esc => match app.tab {
             Tab::Library if matches!(app.focus, Focus::Candidates) => app.close_candidates(),
             Tab::Library if app.detail.is_some() => app.close_detail(),
+            Tab::Library if !app.library_query.is_empty() => app.clear_library_filter(),
             Tab::Add => app.focus = Focus::AddSearchInput,
             Tab::Profiles if app.profile_detail.is_some() => app.close_profile_detail(),
             _ => {}
@@ -155,6 +192,7 @@ async fn handle_key(app: &mut App, code: KeyCode, roots: &LibraryRoots) {
             }
             Tab::Profiles => app.open_profile_detail(),
             Tab::Stuck => app.jump_to_stuck_target().await,
+            Tab::Calendar => app.jump_to_calendar_entry().await,
             _ => {}
         },
         KeyCode::Left | KeyCode::Right if matches!(app.tab, Tab::Stuck) => {
@@ -182,6 +220,16 @@ async fn handle_key(app: &mut App, code: KeyCode, roots: &LibraryRoots) {
         KeyCode::Char('r') if matches!(app.tab, Tab::Review) => {
             app.reject_selected_review().await;
         }
+        KeyCode::Char('/') if matches!(app.tab, Tab::Library) && app.detail.is_none() => {
+            app.start_library_filter();
+        }
+        KeyCode::Char('n')
+            if matches!(app.tab, Tab::Library)
+                && app.detail.is_some()
+                && !matches!(app.focus, Focus::Candidates) =>
+        {
+            app.select_next_missing_episode();
+        }
         KeyCode::Char('s') if matches!(app.tab, Tab::Library) && app.detail.is_some() => {
             app.search_now_selected().await;
         }
@@ -211,9 +259,20 @@ async fn handle_key(app: &mut App, code: KeyCode, roots: &LibraryRoots) {
     }
 }
 
-fn cycle_tab(app: &mut App) {
+fn tab_offset(current: Tab, delta: i32) -> Tab {
-    let idx = Tab::ALL.iter().position(|t| *t == app.tab).unwrap_or(0);
+    let idx = Tab::ALL.iter().position(|t| *t == current).unwrap_or(0) as i32;
-    app.tab = Tab::ALL[(idx + 1) % Tab::ALL.len()];
+    let len = Tab::ALL.len() as i32;
+    Tab::ALL[(idx + delta).rem_euclid(len) as usize]
+}
+
+fn switch_tab(app: &mut App, tab: Tab) {
+    if app.tab == tab {
+        return;
+    }
+    if matches!(app.focus, Focus::Candidates) {
+        app.close_candidates();
+    }
+    app.tab = tab;
     app.detail = None;
     app.profile_detail = None;
     app.profile_weight_state.select(None);

breadarr-tui/src/ui.rs

@@ -5,6 +5,7 @@ use ratatui::widgets::{Block, Borders, Clear, List, ListItem, Paragraph, Tabs};
 use ratatui::Frame;
 
 use crate::app::{App, Focus, StuckSection, Tab};
+use breadarr_shared::dto::CycleInfo;
 
 // Shared color palette — kept to these meanings so a color never has to be
 // second-guessed at a glance:
@@ -58,6 +59,7 @@ fn context_keybindings(app: &App) -> Vec<(&'static str, &'static str)> {
         Tab::Library if app.detail.is_some() => vec![
             ("Esc", "back"),
             ("s", "search now"),
+            ("n", "next missing"),
             ("m", "monitor show"),
             ("e", "monitor episode"),
             ("S", "monitor season"),
@@ -65,8 +67,12 @@ fn context_keybindings(app: &App) -> Vec<(&'static str, &'static str)> {
             ("d", "delete file"),
             ("c", "pick release"),
         ],
+        Tab::Library if matches!(app.focus, Focus::LibraryFilterInput) => {
+            vec![("Enter", "keep filter"), ("Esc", "clear filter")]
+        }
         Tab::Library => vec![
             ("Enter", "open detail"),
+            ("/", "filter title"),
             ("f", "cycle filter"),
             ("o", "cycle sort"),
             ("m", "monitor toggle"),
@@ -81,13 +87,17 @@ fn context_keybindings(app: &App) -> Vec<(&'static str, &'static str)> {
         }
         Tab::Profiles => vec![("Enter", "open profile")],
         Tab::Stuck => vec![("Left/Right", "switch section"), ("Enter", "jump to show")],
+        Tab::Calendar => vec![("Enter", "jump to show")],
         _ => vec![],
     }
 }
 
 const GLOBAL_KEYS: &[(&str, &str)] = &[
-    ("Tab", "switch tab"),
+    ("Tab/S-Tab", "switch tab"),
+    ("1-8", "jump tab"),
     ("j/k", "move"),
+    ("g/G", "first/last"),
+    ("PgUp/PgDn", "page"),
     ("?", "help"),
     ("R", "refresh now"),
     ("q", "quit"),
@@ -128,7 +138,7 @@ fn draw_help_overlay(frame: &mut Frame, area: Rect, app: &App) {
         ])
     }));
 
-    let popup = centered_rect(50, lines.len() as u16 + 2, area);
+    let popup = centered_rect(52, lines.len() as u16 + 2, area);
     frame.render_widget(Clear, popup);
     let paragraph = Paragraph::new(lines).block(
         Block::default()
@@ -138,8 +148,47 @@ fn draw_help_overlay(frame: &mut Frame, area: Rect, app: &App) {
     frame.render_widget(paragraph, popup);
 }
 
+fn tab_label(tab: Tab, app: &App) -> String {
+    match tab {
+        Tab::Library => {
+            let missing: i64 = app.media_items.iter().map(|m| m.missing_count).sum();
+            if missing > 0 {
+                format!("Library ({missing})")
+            } else {
+                tab.title().to_string()
+            }
+        }
+        Tab::Review => {
+            let n = if matches!(app.tab, Tab::Review) {
+                app.review_items.len()
+            } else {
+                app.review_count
+            };
+            if n > 0 {
+                format!("Review ({n})")
+            } else {
+                tab.title().to_string()
+            }
+        }
+        Tab::Stuck => {
+            let n = app.stuck.as_ref().map_or(app.stuck_count, |r| {
+                r.stalled_grabs.len() + r.maxed_out_search_targets.len()
+            });
+            if n > 0 {
+                format!("Stuck ({n})")
+            } else {
+                tab.title().to_string()
+            }
+        }
+        _ => tab.title().to_string(),
+    }
+}
+
 fn draw_tabs(frame: &mut Frame, area: Rect, app: &App) {
-    let titles: Vec<Line> = Tab::ALL.iter().map(|t| Line::from(t.title())).collect();
+    let titles: Vec<Line> = Tab::ALL
+        .iter()
+        .map(|t| Line::from(tab_label(*t, app)))
+        .collect();
     let selected = Tab::ALL.iter().position(|t| *t == app.tab).unwrap_or(0);
 
     let (daemon_label, daemon_color) = match (&app.daemon_up, &app.health) {
@@ -190,6 +239,11 @@ fn draw_library(frame: &mut Frame, area: Rect, app: &App) {
             return;
         }
 
+        if detail.kind == "movie" || detail.episodes.is_empty() {
+            draw_movie_detail(frame, area, app, detail);
+            return;
+        }
+
         let items: Vec<ListItem> = detail
             .episodes
             .iter()
@@ -286,9 +340,19 @@ fn draw_library(frame: &mut Frame, area: Rect, app: &App) {
             ]))
         })
         .collect();
+    let search = if app.library_query.is_empty() {
+        String::new()
+    } else {
+        format!("  search: {}", app.library_query)
+    };
+    let filter_caret = if matches!(app.focus, Focus::LibraryFilterInput) {
+        "▋"
+    } else {
+        ""
+    };
     let list = List::new(items)
         .block(Block::default().borders(Borders::ALL).title(format!(
-            "Monitored Shows ({}/{}) — filter: {}  sort: {}",
+            "Library ({}/{}) — filter: {}  sort: {}{search}{filter_caret}",
             app.library_view.len(),
             app.media_items.len(),
             app.library_filter.label(),
@@ -299,6 +363,61 @@ fn draw_library(frame: &mut Frame, area: Rect, app: &App) {
     frame.render_stateful_widget(list, area, &mut state);
 }
 
+fn draw_movie_detail(
+    frame: &mut Frame,
+    area: Rect,
+    app: &App,
+    detail: &breadarr_shared::dto::MediaItemDetail,
+) {
+    let confirm = if app.confirm_delete {
+        " — x AGAIN TO DELETE"
+    } else if app.confirm_delete_file {
+        " — d AGAIN TO DELETE FILE"
+    } else {
+        ""
+    };
+    let monitor_label = if detail.monitored {
+        "monitored"
+    } else {
+        "unmonitored"
+    };
+    let have = app
+        .media_items
+        .iter()
+        .find(|m| m.id == detail.id)
+        .map(|m| m.missing_count == 0);
+    let file_line = match have {
+        Some(true) => ("file on disk", Color::Green),
+        Some(false) => ("missing", Color::Yellow),
+        None => ("file status unknown", Color::DarkGray),
+    };
+    let kind = if detail.kind == "movie" {
+        "Movie"
+    } else {
+        "Series"
+    };
+    let year = detail
+        .year
+        .map(|y| y.to_string())
+        .unwrap_or_else(|| "—".into());
+    let lines = vec![
+        Line::from(Span::styled(
+            format!("{} ({year})", detail.title),
+            Style::default().add_modifier(Modifier::BOLD),
+        )),
+        Line::from(""),
+        Line::from(format!("{kind}  ·  {monitor_label}")),
+        Line::from(Span::styled(file_line.0, Style::default().fg(file_line.1))),
+        Line::from(format!("root: {}", detail.root_folder)),
+        Line::from(""),
+        Line::from("s search now   c pick release   m monitor   d delete file   x remove"),
+    ];
+    let paragraph = Paragraph::new(lines).block(Block::default().borders(Borders::ALL).title(
+        format!("{} ({year}) [{monitor_label}]{confirm}", detail.title),
+    ));
+    frame.render_widget(paragraph, area);
+}
+
 /// Manual release picker — candidates for whatever episode/movie was
 /// selected when `c` was pressed, scored (or gate-rejected with a reason)
 /// exactly like the automatic search pipeline would see them.
@@ -330,15 +449,19 @@ fn draw_candidates(frame: &mut Frame, area: Rect, app: &App, media_title: &str)
                 if c.is_season_pack { " [PACK]" } else { "" },
                 if c.is_repack { " [REPACK]" } else { "" },
             );
-            let verdict = match (c.score, &c.rejected_reason) {
+            let (verdict, color) = match (c.score, &c.rejected_reason) {
-                (Some(score), _) => format!("score {score:.1}"),
+                (Some(score), _) if score >= 8.0 => (format!("score {score:.1}"), Color::Green),
-                (None, Some(reason)) => format!("REJECTED: {reason}"),
+                (Some(score), _) => (format!("score {score:.1}"), Color::Yellow),
-                (None, None) => "unscored".to_string(),
+                (None, Some(reason)) => (format!("REJECTED: {reason}"), Color::Red),
+                (None, None) => ("unscored".to_string(), Color::DarkGray),
             };
-            ListItem::new(format!(
+            ListItem::new(Line::from(vec![
-                "[{}] {} — {seeders} seeders, {size}{flags} — {verdict}",
+                Span::raw(format!(
+                    "[{}] {} — {seeders} seeders, {size}{flags} — ",
                     c.source_name, c.raw_title
-            ))
+                )),
+                Span::styled(verdict, Style::default().fg(color)),
+            ]))
         })
         .collect();
     let list = List::new(items)
@@ -350,18 +473,32 @@ fn draw_candidates(frame: &mut Frame, area: Rect, app: &App, media_title: &str)
     frame.render_stateful_widget(list, area, &mut state);
 }
 
+fn history_status_color(status: &str) -> Color {
+    match status {
+        "imported" => Color::Green,
+        "grabbed" => Color::Yellow,
+        "failed" => Color::Red,
+        _ => Color::DarkGray,
+    }
+}
+
 fn draw_history(frame: &mut Frame, area: Rect, app: &App) {
     let items: Vec<ListItem> = app
         .releases
         .iter()
         .map(|r| {
-            ListItem::new(format!(
+            ListItem::new(Line::from(vec![
-                "[{}] {} — {} (score {:.1})",
+                Span::styled(
-                r.status,
+                    format!("[{}]", r.status),
+                    Style::default().fg(history_status_color(&r.status)),
+                ),
+                Span::raw(format!(
+                    " {} — {} (score {:.1})",
                     r.media_title,
                     r.raw_title,
                     r.score.unwrap_or(0.0)
-            ))
+                )),
+            ]))
         })
         .collect();
     let list = List::new(items)
@@ -384,7 +521,10 @@ fn draw_review(frame: &mut Frame, area: Rect, app: &App) {
                 Color::Green
             };
             ListItem::new(Line::from(vec![
-                Span::styled(format!("({:.0}%)", r.confidence * 100.0), Style::default().fg(color)),
+                Span::styled(
+                    format!("({:.0}%)", r.confidence * 100.0),
+                    Style::default().fg(color),
+                ),
                 Span::raw(format!(
                     " {} -> {}",
                     r.raw_release_title,
@@ -439,7 +579,10 @@ fn draw_stuck(frame: &mut Frame, area: Rect, app: &App) {
         .iter()
         .map(|g| {
             ListItem::new(Line::from(Span::styled(
-                format!("{} — {} (grabbed {})", g.media_title, g.raw_title, g.grabbed_at),
+                format!(
+                    "{} — {} (grabbed {})",
+                    g.media_title, g.raw_title, g.grabbed_at
+                ),
                 Style::default().fg(Color::Yellow),
             )))
         })
@@ -519,12 +662,21 @@ fn draw_add(frame: &mut Frame, area: Rect, app: &App) {
         .add_results
         .iter()
         .map(|r| {
-            ListItem::new(format!(
+            let kind_color = match r.kind {
-                "[{}] {} ({})",
+                crate::app::AddKind::Movie => Color::Magenta,
-                r.kind.label(),
+                crate::app::AddKind::Series => Color::Blue,
+            };
+            ListItem::new(Line::from(vec![
+                Span::styled(
+                    format!("[{}] ", r.kind.label()),
+                    Style::default().fg(kind_color),
+                ),
+                Span::raw(format!(
+                    "{} ({})",
                     r.result.title,
                     r.result.year.map(|y| y.to_string()).unwrap_or_default()
-            ))
+                )),
+            ]))
         })
         .collect();
     let list = List::new(items)
@@ -539,36 +691,43 @@ fn draw_add(frame: &mut Frame, area: Rect, app: &App) {
 }
 
 /// What's aired recently or airs soon (a week back, three weeks forward —
-/// see `calendar::DAYS_PAST`/`DAYS_FUTURE` server-side). Read-only, no
+/// see `calendar::DAYS_PAST`/`DAYS_FUTURE` server-side). Selectable —
-/// selection — a lookahead view, not something acted on directly here.
+/// Enter jumps to that episode in Library.
 fn draw_calendar(frame: &mut Frame, area: Rect, app: &App) {
     let today = chrono::Local::now().date_naive().to_string();
     let items: Vec<ListItem> = app
         .calendar
         .iter()
         .map(|e| {
-            let status = if e.has_file {
+            let (status, color) = if e.has_file {
-                "✓"
+                ("✓", Color::Green)
             } else if !e.monitored {
-                "-"
+                ("-", Color::DarkGray)
             } else if e.air_date.as_str() > today.as_str() {
-                "…"
+                ("…", Color::Yellow)
             } else {
-                "!" // aired, monitored, still missing
+                ("!", Color::Red)
             };
+            let today_mark = if e.air_date == today { "  today" } else { "" };
             let title = e.title.as_deref().unwrap_or("");
-            ListItem::new(format!(
+            ListItem::new(Line::from(vec![
-                "{status} {}  {}  S{:02}E{:02}  {title}",
+                Span::styled(status, Style::default().fg(color)),
+                Span::raw(format!(
+                    " {}  {}  S{:02}E{:02}  {title}{today_mark}",
                     e.air_date, e.media_title, e.season_number, e.episode_number
-            ))
+                )),
+            ]))
         })
         .collect();
-    let list = List::new(items).block(
+    let list = List::new(items)
+        .block(
             Block::default()
                 .borders(Borders::ALL)
-            .title("Calendar — ✓ have it  … upcoming  ! aired but missing"),
+                .title("Calendar — Enter: jump  ✓ have it  … upcoming  ! aired but missing"),
-    );
+        )
-    frame.render_widget(list, area);
+        .highlight_style(Style::default().add_modifier(Modifier::REVERSED));
+    let mut state = app.calendar_state.clone();
+    frame.render_stateful_widget(list, area, &mut state);
 }
 
 /// Read-only report on `media_file_probe` state: corruption, under-quality,
@@ -588,7 +747,7 @@ fn draw_library_health(frame: &mut Frame, area: Rect, app: &App) {
 
     let chunks = Layout::default()
         .direction(Direction::Vertical)
-        .constraints([Constraint::Length(5), Constraint::Min(3)])
+        .constraints([Constraint::Length(6), Constraint::Min(3)])
         .split(area);
 
     let s = &report.summary;
@@ -598,8 +757,13 @@ fn draw_library_health(frame: &mut Frame, area: Rect, app: &App) {
         .map(|c| format!("{}={}", c.codec, c.count))
         .collect::<Vec<_>>()
         .join(", ");
+    let cycles = app
+        .health
+        .as_ref()
+        .map(cycle_summary_line)
+        .unwrap_or_default();
     let summary_text = format!(
-        "{} files, {:.1} GB total, {} probed — resolution: SD={} 720p={} 1080p={} 4K={} \
+        "{cycles}\n{} files, {:.1} GB total, {} probed — resolution: SD={} 720p={} 1080p={} 4K={} \
          — subtitles: {:.0}% — codecs: {codec_summary}",
         s.total_files,
         s.total_size_bytes as f64 / 1_073_741_824.0,
@@ -668,12 +832,38 @@ fn draw_library_health(frame: &mut Frame, area: Rect, app: &App) {
         items.push(ListItem::new("Nothing flagged — library looks clean."));
     }
 
-    let list = List::new(items).block(
+    let list = List::new(items)
+        .block(
             Block::default()
                 .borders(Borders::ALL)
                 .title("Flagged files"),
-    );
+        )
-    frame.render_widget(list, chunks[1]);
+        .highlight_style(Style::default().add_modifier(Modifier::REVERSED));
+    let mut state = app.health_state.clone();
+    frame.render_stateful_widget(list, chunks[1], &mut state);
+}
+
+fn cycle_bit(name: &str, info: &Option<CycleInfo>) -> String {
+    match info {
+        Some(c) if c.ok => format!("{name}: ok"),
+        Some(_) => format!("{name}: FAIL"),
+        None => format!("{name}: —"),
+    }
+}
+
+fn cycle_summary_line(h: &breadarr_shared::dto::HealthDetail) -> String {
+    format!(
+        "{}  |  {}  |  {}  |  {}{}",
+        cycle_bit("grab", &h.last_grab_cycle),
+        cycle_bit("import", &h.last_import_cycle),
+        cycle_bit("search", &h.last_search_cycle),
+        cycle_bit("upgrade", &h.last_upgrade_cycle),
+        if h.search_halted {
+            "  |  search HALTED"
+        } else {
+            ""
+        }
+    )
 }
 
 /// Quality-profile weight editing — list of profiles, then (once one is

breadarrd/src/main.rs

@@ -355,6 +355,26 @@ async fn background_loop(
         ) {
             error!(error = %e, "failed to register tpb source row");
         }
+        if let Err(e) = conn.execute(
+            "INSERT OR IGNORE INTO source (id, name, kind, base_url, poll_interval_secs, enabled)
+             VALUES (4, 'torrents-csv', 'scrape', ?1, ?2, 1)",
+            rusqlite::params![
+                config.sources.torrents_csv_url,
+                config.sources.search_poll_interval_secs
+            ],
+        ) {
+            error!(error = %e, "failed to register torrents-csv source row");
+        }
+        if let Err(e) = conn.execute(
+            "INSERT OR IGNORE INTO source (id, name, kind, base_url, poll_interval_secs, enabled)
+             VALUES (5, 'yts', 'scrape', ?1, ?2, 1)",
+            rusqlite::params![
+                config.sources.yts_api_url,
+                config.sources.search_poll_interval_secs
+            ],
+        ) {
+            error!(error = %e, "failed to register yts source row");
+        }
     }
 
     // A transient failure here (network blip during the one-time model
@@ -377,6 +397,21 @@ async fn background_loop(
     let scrape_source =
         sources::scrape::ScrapeSource::new(config.sources.torrent_1337x_mirrors.clone());
     let tpb_source = sources::tpb::TpbSource::new(config.sources.tpb_api_url.clone());
+    let torrents_csv_source =
+        sources::torrents_csv::TorrentsCsvSource::new(config.sources.torrents_csv_url.clone());
+    let yts_source = sources::yts::YtsSource::new(config.sources.yts_api_url.clone());
+    let search_sources = scheduler::SearchSources {
+        tpb: &tpb_source,
+        tpb_id: 3,
+        torrents_csv: &torrents_csv_source,
+        torrents_csv_id: 4,
+        yts: &yts_source,
+        yts_id: 5,
+        scrape: &scrape_source,
+        scrape_id: 2,
+        nyaa_search: &nyaa_source,
+        nyaa_id: 1,
+    };
 
     let mut grab_ticker = tokio::time::interval(std::time::Duration::from_secs(
         config.sources.grab_poll_interval_secs,
@@ -501,9 +536,7 @@ async fn background_loop(
                     let conn = conn.lock().await;
                     scheduler::run_search_cycle(
                         &conn,
-                        &tpb_source, 3,
+                        &search_sources,
-                        &scrape_source, 2,
-                        &nyaa_source, 1,
                         &mut title_matcher,
                         &qbit, &config.qbit.category,
                         config.sources.search_budget_per_cycle,
@@ -562,9 +595,7 @@ async fn background_loop(
                     let conn = conn.lock().await;
                     scheduler::run_upgrade_cycle(
                         &conn,
-                        &tpb_source, 3,
+                        &search_sources,
-                        &scrape_source, 2,
-                        &nyaa_source, 1,
                         &mut title_matcher,
                         &qbit, &config.qbit.category,
                         config.sources.upgrade_budget_per_cycle,
@@ -679,9 +710,7 @@ async fn background_loop(
                                 Ok(targets) => scheduler::execute_search_targets(
                                     &conn,
                                     &targets,
-                                    &tpb_source, 3,
+                                    &search_sources,
-                                    &scrape_source, 2,
-                                    &nyaa_source, 1,
                                     &mut title_matcher,
                                     &qbit, &config.qbit.category,
                                 ).await,
@@ -705,9 +734,7 @@ async fn background_loop(
                                 &conn,
                                 media_item_id,
                                 episode_id,
-                                &tpb_source, 3,
+                                &search_sources,
-                                &scrape_source, 2,
-                                &nyaa_source, 1,
                             ).await
                         };
                         let _ = reply.send(result);
@@ -1133,6 +1160,22 @@ async fn debug_search_show(config: &Config, title: &str) -> Result<()> {
             config.sources.search_poll_interval_secs
         ],
     )?;
+    conn.execute(
+        "INSERT OR IGNORE INTO source (id, name, kind, base_url, poll_interval_secs, enabled)
+         VALUES (4, 'torrents-csv', 'scrape', ?1, ?2, 1)",
+        rusqlite::params![
+            config.sources.torrents_csv_url,
+            config.sources.search_poll_interval_secs
+        ],
+    )?;
+    conn.execute(
+        "INSERT OR IGNORE INTO source (id, name, kind, base_url, poll_interval_secs, enabled)
+         VALUES (5, 'yts', 'scrape', ?1, ?2, 1)",
+        rusqlite::params![
+            config.sources.yts_api_url,
+            config.sources.search_poll_interval_secs
+        ],
+    )?;
 
     let qbit = QbitClient::new(config.qbit.base_url.clone())?;
     if !config.qbit.username.is_empty() {
@@ -1144,6 +1187,21 @@ async fn debug_search_show(config: &Config, title: &str) -> Result<()> {
     let scrape_source =
         sources::scrape::ScrapeSource::new(config.sources.torrent_1337x_mirrors.clone());
     let nyaa_source = sources::rss::RssSource::new(config.sources.nyaa_rss_url.clone());
+    let torrents_csv_source =
+        sources::torrents_csv::TorrentsCsvSource::new(config.sources.torrents_csv_url.clone());
+    let yts_source = sources::yts::YtsSource::new(config.sources.yts_api_url.clone());
+    let search_sources = scheduler::SearchSources {
+        tpb: &tpb_source,
+        tpb_id: 3,
+        torrents_csv: &torrents_csv_source,
+        torrents_csv_id: 4,
+        yts: &yts_source,
+        yts_id: 5,
+        scrape: &scrape_source,
+        scrape_id: 2,
+        nyaa_search: &nyaa_source,
+        nyaa_id: 1,
+    };
 
     let targets = scheduler::enumerate_search_targets_for_media_item(&conn, media_item_id)?;
     println!("{} missing episode(s)/movie for {title:?}", targets.len());
@@ -1151,12 +1209,7 @@ async fn debug_search_show(config: &Config, title: &str) -> Result<()> {
     let stats = scheduler::execute_search_targets(
         &conn,
         &targets,
-        &tpb_source,
+        &search_sources,
-        3,
-        &scrape_source,
-        2,
-        &nyaa_source,
-        1,
         &mut title_matcher,
         &qbit,
         &config.qbit.category,

breadarrd/src/scheduler.rs

@@ -652,7 +652,8 @@ async fn process_item(
         return Ok(ProcessOutcome::QueuedForReview);
     }
 
-    let release_score = scoring::score(&parsed, item.seeders.unwrap_or(0), parsed.has_hdr, &profile);
+    let release_score =
+        scoring::score(&parsed, item.seeders.unwrap_or(0), parsed.has_hdr, &profile);
     let existing_best = match (episode_id, season_pack_number) {
         (Some(eid), _) => best_existing_score(conn, eid)?,
         (None, Some(season)) => best_existing_season_pack_score(conn, media_item.id, season)?,
@@ -911,8 +912,8 @@ pub async fn run_grab_cycle(
     Ok(stats)
 }
 
-// --- Search-driven acquisition (1337x for general TV/movies, nyaa search
+// --- Search-driven acquisition (TPB + YTS/csv/1337x fallbacks for
-// for anime movies) ---
+// general TV/movies, nyaa search for anime movies) ---
 //
 // Unlike the feed-based path above, there's no natural stream of "new"
 // items to dedup against — the recurring cost here is the *search itself*,
@@ -921,18 +922,12 @@ pub async fn run_grab_cycle(
 // single cycle; cadence backs off exponentially (6h, 12h, 24h, 48h, 96h,
 // capped at a week) the more times it's been searched without success.
 
-/// Other general-content sources considered and rejected (live-tested
+/// General-content fallbacks live-tested 2026-08-16 (TPB/apibay was
-/// 2026-07-12, not just assumed) before landing on TPB as primary:
+/// timing out; every configured 1337x mirror returned Cloudflare 521):
-/// - **YTS** (`yts.mx`): DNS doesn't resolve at all. Every known mirror
+/// - **torrents-csv** and **YTS** (`yts.lt` API — `yts.mx` still does not
-///   (`yts.am`, `yts.ag`, `yts.lt`, `yts.pe`) either 301s in a loop or drops
+///   resolve) are JSON hash-to-magnet sources, same grab shape as TPB.
-///   the query and lands on a bare homepage. The whole mirror network looks
+/// - **EZTV**'s JSON API is up but IMDb-id only; name search is a
-///   dead, not just one domain — re-check before assuming a fix is quick.
+///   Cloudflare challenge. Not wired — breadarr has TMDB/TVDB, not IMDb.
-/// - **EZTV** (`eztv.re`): redirects to `eztvx.to`, which fails to connect
-///   outright (TLS/connection error, not a slow response). Also
-///   Cloudflare-fronted, so even if connectivity is restored it carries the
-///   same risk profile 1337x does.
-///   If revisiting either, re-verify connectivity first — this isn't a
-///   permanent architectural decision, just what was true when checked.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
 enum SearchRoute {
     /// Primary general-content (movies + non-anime TV) route — a JSON API,
@@ -947,6 +942,22 @@ enum SearchRoute {
     NyaaSearch,
 }
 
+/// The search-driven sources plus their `source` table ids — one bundle
+/// so `execute_search_targets` / `fetch_candidates` / the cycle runners
+/// don't each grow another four arguments every time a fallback is added.
+pub struct SearchSources<'a> {
+    pub tpb: &'a sources::tpb::TpbSource,
+    pub tpb_id: i64,
+    pub torrents_csv: &'a sources::torrents_csv::TorrentsCsvSource,
+    pub torrents_csv_id: i64,
+    pub yts: &'a sources::yts::YtsSource,
+    pub yts_id: i64,
+    pub scrape: &'a sources::scrape::ScrapeSource,
+    pub scrape_id: i64,
+    pub nyaa_search: &'a sources::rss::RssSource,
+    pub nyaa_id: i64,
+}
+
 #[derive(Debug, Clone, PartialEq)]
 pub struct SearchTarget {
     media_item_id: i64,
@@ -1565,35 +1576,16 @@ pub struct SearchCycleStats {
 /// single broad query can dump into the review queue.
 const MAX_RESULTS_PER_SEARCH: usize = 15;
 
-#[allow(clippy::too_many_arguments)]
 pub async fn run_search_cycle(
     conn: &Connection,
-    tpb: &sources::tpb::TpbSource,
+    sources: &SearchSources<'_>,
-    tpb_source_id: i64,
-    scrape: &sources::scrape::ScrapeSource,
-    scrape_source_id: i64,
-    nyaa_search: &sources::rss::RssSource,
-    nyaa_source_id: i64,
     matcher: &mut TitleMatcher,
     qbit: &QbitClient,
     qbit_category: &str,
     budget: usize,
 ) -> Result<SearchCycleStats> {
     let targets = enumerate_search_targets(conn, budget)?;
-    execute_search_targets(
+    execute_search_targets(conn, &targets, sources, matcher, qbit, qbit_category).await
-        conn,
-        &targets,
-        tpb,
-        tpb_source_id,
-        scrape,
-        scrape_source_id,
-        nyaa_search,
-        nyaa_source_id,
-        matcher,
-        qbit,
-        qbit_category,
-    )
-    .await
 }
 
 /// Upgrade-search counterpart to `run_search_cycle`: same execution engine
@@ -1601,15 +1593,9 @@ pub async fn run_search_cycle(
 /// ones. `min_gain` is threaded onto every target via
 /// `enumerate_upgrade_targets`, which is what routes `process_item` into
 /// its upgrade-eligibility path instead of the normal missing-content one.
-#[allow(clippy::too_many_arguments)]
 pub async fn run_upgrade_cycle(
     conn: &Connection,
-    tpb: &sources::tpb::TpbSource,
+    sources: &SearchSources<'_>,
-    tpb_source_id: i64,
-    scrape: &sources::scrape::ScrapeSource,
-    scrape_source_id: i64,
-    nyaa_search: &sources::rss::RssSource,
-    nyaa_source_id: i64,
     matcher: &mut TitleMatcher,
     qbit: &QbitClient,
     qbit_category: &str,
@@ -1617,20 +1603,7 @@ pub async fn run_upgrade_cycle(
     min_gain: f32,
 ) -> Result<SearchCycleStats> {
     let targets = enumerate_upgrade_targets(conn, budget, min_gain)?;
-    execute_search_targets(
+    execute_search_targets(conn, &targets, sources, matcher, qbit, qbit_category).await
-        conn,
-        &targets,
-        tpb,
-        tpb_source_id,
-        scrape,
-        scrape_source_id,
-        nyaa_search,
-        nyaa_source_id,
-        matcher,
-        qbit,
-        qbit_category,
-    )
-    .await
 }
 
 /// Every currently-missing episode/movie for one specific `media_item`,
@@ -1728,16 +1701,10 @@ pub fn find_media_item_id_by_title(conn: &Connection, title: &str) -> Result<Opt
     .map_err(Into::into)
 }
 
-#[allow(clippy::too_many_arguments)]
 pub async fn execute_search_targets(
     conn: &Connection,
     targets: &[SearchTarget],
-    tpb: &sources::tpb::TpbSource,
+    sources: &SearchSources<'_>,
-    tpb_source_id: i64,
-    scrape: &sources::scrape::ScrapeSource,
-    scrape_source_id: i64,
-    nyaa_search: &sources::rss::RssSource,
-    nyaa_source_id: i64,
     matcher: &mut TitleMatcher,
     qbit: &QbitClient,
     qbit_category: &str,
@@ -1753,8 +1720,8 @@ pub async fn execute_search_targets(
     // without this, a 10-episode backlog fires 10 indistinguishable
     // requests at a single-domain API with no mirror fallback, which is
     // exactly the kind of pattern that gets a source rate-limited. Caches
-    // which source actually answered too (the TPB→1337x fallback can mean
+    // which source actually answered too (the TPB → YTS/csv/1337x chain
-    // two different targets with the same query string were served by two
+    // can mean two targets with the same query were served by two
     // different sources), so a cache hit still attributes dedup/grab
     // records to the right source id.
     let mut query_cache: std::collections::HashMap<
@@ -1763,12 +1730,6 @@ pub async fn execute_search_targets(
     > = std::collections::HashMap::new();
 
     for (i, target) in targets.iter().enumerate() {
-        let (primary, primary_id): (&dyn ReleaseSource, i64) = match target.route {
-            SearchRoute::Tpb => (tpb, tpb_source_id),
-            SearchRoute::X1337 => (scrape, scrape_source_id),
-            SearchRoute::NyaaSearch => (nyaa_search, nyaa_source_id),
-        };
-
         let cache_key = (target.route, target.query.clone());
         let (items, source_id) =
             if let Some((cached_items, cached_source_id)) = query_cache.get(&cache_key) {
@@ -1779,34 +1740,11 @@ pub async fn execute_search_targets(
                     tokio::time::sleep(std::time::Duration::from_secs(jitter_secs)).await;
                 }
 
-                let primary_result = primary.fetch(Some(&target.query)).await;
+                match fetch_for_target(target, sources).await {
-                // TPB is the primary route for general content, but a fetch
+                    Ok(pair) => {
-                // *failure* there (not just "no relevant results") falls back
-                // to 1337x for the same query before giving up — keeps the
-                // mirror-rotation/cooldown machinery built for 1337x as real
-                // resilience rather than dead code, just no longer the first
-                // choice given TPB's better precision. `result_source_id`
-                // tracks which source actually produced whatever we end up
-                // with, since dedup (`is_seen`/`mark_seen`) and grab records
-                // are keyed by source id — attributing a 1337x-sourced guid to
-                // TPB's source id would silently break dedup between the two.
-                let (fetch_result, result_source_id) = match primary_result {
-                    Err(e) if matches!(target.route, SearchRoute::Tpb) => {
-                        tracing::warn!(
-                            query = %target.query,
-                            error = %e,
-                            "TPB search failed, falling back to 1337x"
-                        );
-                        (scrape.fetch(Some(&target.query)).await, scrape_source_id)
-                    }
-                    other => (other, primary_id),
-                };
-
-                match fetch_result {
-                    Ok(items) => {
                         consecutive_fetch_errors = 0;
-                        query_cache.insert(cache_key, (items.clone(), result_source_id));
+                        query_cache.insert(cache_key, pair.clone());
-                        (items, result_source_id)
+                        pair
                     }
                     Err(e) => {
                         consecutive_fetch_errors += 1;
@@ -1937,14 +1875,96 @@ pub async fn execute_search_targets(
     Ok(stats)
 }
 
-fn source_route_name(route: SearchRoute) -> &'static str {
+fn source_name_for_id(id: i64) -> &'static str {
-    match route {
+    match id {
-        SearchRoute::Tpb => "tpb",
+        1 => "nyaa",
-        SearchRoute::X1337 => "1337x",
+        2 => "1337x",
-        SearchRoute::NyaaSearch => "nyaa",
+        3 => "tpb",
+        4 => "torrents-csv",
+        5 => "yts",
+        _ => "unknown",
     }
 }
 
+/// TPB first; on failure *or* empty results, walk the supplement chain
+/// (torrents-csv for everything, YTS for movies, 1337x last). Empty is
+/// treated as "try the next one" so a live-but-empty apibay doesn't hide
+/// a title that YTS/csv actually has. Dedup/grabs use whichever source
+/// actually answered.
+async fn fetch_for_target(
+    target: &SearchTarget,
+    sources: &SearchSources<'_>,
+) -> Result<(Vec<RawReleaseItem>, i64)> {
+    match target.route {
+        SearchRoute::NyaaSearch => Ok((
+            sources.nyaa_search.fetch(Some(&target.query)).await?,
+            sources.nyaa_id,
+        )),
+        SearchRoute::X1337 => Ok((
+            sources.scrape.fetch(Some(&target.query)).await?,
+            sources.scrape_id,
+        )),
+        SearchRoute::Tpb => fetch_general_content(target, sources).await,
+    }
+}
+
+async fn fetch_general_content(
+    target: &SearchTarget,
+    sources: &SearchSources<'_>,
+) -> Result<(Vec<RawReleaseItem>, i64)> {
+    let is_movie = target.episode_id.is_none();
+    let mut attempts: Vec<(&dyn ReleaseSource, i64, &'static str)> =
+        vec![(sources.tpb, sources.tpb_id, "tpb")];
+    if is_movie {
+        attempts.push((sources.yts, sources.yts_id, "yts"));
+    }
+    attempts.push((
+        sources.torrents_csv,
+        sources.torrents_csv_id,
+        "torrents-csv",
+    ));
+    attempts.push((sources.scrape, sources.scrape_id, "1337x"));
+
+    let mut last_err: Option<anyhow::Error> = None;
+    let mut any_ok = false;
+    for (src, id, name) in attempts {
+        match src.fetch(Some(&target.query)).await {
+            Ok(items) if !items.is_empty() => {
+                if name != "tpb" {
+                    tracing::info!(
+                        query = %target.query,
+                        source = name,
+                        n = items.len(),
+                        "search fallback produced results"
+                    );
+                }
+                return Ok((items, id));
+            }
+            Ok(_) => {
+                any_ok = true;
+                tracing::debug!(
+                    query = %target.query,
+                    source = name,
+                    "search source returned no results"
+                );
+            }
+            Err(e) => {
+                tracing::warn!(
+                    query = %target.query,
+                    source = name,
+                    error = %e,
+                    "search source failed"
+                );
+                last_err = Some(e);
+            }
+        }
+    }
+    if any_ok {
+        return Ok((Vec::new(), sources.tpb_id));
+    }
+    Err(last_err.unwrap_or_else(|| anyhow::anyhow!("all general-content sources failed")))
+}
+
 /// Fetches and scores (or gate-rejects) candidates for one search target —
 /// the same evaluation `execute_search_targets` does automatically, minus
 /// the grab decision, surfaced instead for a human to choose from. Used by
@@ -1954,29 +1974,18 @@ fn source_route_name(route: SearchRoute) -> &'static str {
 /// episode/movie right now (already owned, unmonitored, or mid-grab) —
 /// same "nothing to do" cases `enumerate_search_targets_for_media_item`
 /// already excludes.
-#[allow(clippy::too_many_arguments)]
 pub async fn fetch_candidates(
     conn: &Connection,
     media_item_id: i64,
     episode_id: Option<i64>,
-    tpb: &sources::tpb::TpbSource,
+    sources: &SearchSources<'_>,
-    tpb_source_id: i64,
-    scrape: &sources::scrape::ScrapeSource,
-    scrape_source_id: i64,
-    nyaa_search: &sources::rss::RssSource,
-    nyaa_source_id: i64,
 ) -> Result<Vec<breadarr_shared::dto::ReleaseCandidate>> {
     let targets = enumerate_search_targets_for_media_item(conn, media_item_id)?;
     let Some(target) = targets.into_iter().find(|t| t.episode_id == episode_id) else {
         return Ok(Vec::new());
     };
 
-    let (source, source_id): (&dyn ReleaseSource, i64) = match target.route {
+    let (items, source_id) = fetch_for_target(&target, sources).await?;
-        SearchRoute::Tpb => (tpb, tpb_source_id),
-        SearchRoute::X1337 => (scrape, scrape_source_id),
-        SearchRoute::NyaaSearch => (nyaa_search, nyaa_source_id),
-    };
-    let items = source.fetch(Some(&target.query)).await?;
 
     let media_item = get_media_item(conn, media_item_id)?;
     let anime = match media_item.tvdb_id {
@@ -2029,7 +2038,7 @@ pub async fn fetch_candidates(
             link: item.link.clone(),
             guid: item.guid.clone(),
             source_id,
-            source_name: source_route_name(target.route).to_string(),
+            source_name: source_name_for_id(source_id).to_string(),
             seeders: item.seeders,
             leechers: item.leechers,
             size_bytes: item.size_bytes,

breadarrd/src/sources/mod.rs

@@ -1,6 +1,8 @@
 pub mod rss;
 pub mod scrape;
+pub mod torrents_csv;
 pub mod tpb;
+pub mod yts;
 
 use anyhow::Result;
 use async_trait::async_trait;
@@ -25,6 +27,39 @@ pub trait ReleaseSource {
     async fn fetch(&self, query: Option<&str>) -> Result<Vec<RawReleaseItem>>;
 }
 
+/// Trackers attached to every magnet we synthesize from an info-hash
+/// (TPB, torrents-csv, YTS). Same set the TPB client has used since it
+/// landed — qBittorrent needs *some* announce list or the torrent sits
+/// hash-only until DHT finds peers.
+const MAGNET_TRACKERS: &[&str] = &[
+    "udp://tracker.opentrackr.org:1337/announce",
+    "udp://open.stealth.si:80/announce",
+    "udp://tracker.torrent.eu.org:451/announce",
+    "udp://tracker.openbittorrent.com:6969/announce",
+    "udp://exodus.desync.com:6969/announce",
+];
+
+/// A valid BitTorrent v1 info_hash: 40 hex chars or 32 base32 chars — same
+/// shape `qbit::extract_btih` accepts out of a magnet URI. Shared by every
+/// hash-to-magnet source so a malformed value can't silently produce a
+/// magnet the grab path then fails to parse back.
+pub(crate) fn is_valid_info_hash(hash: &str) -> bool {
+    (hash.len() == 40 && hash.bytes().all(|b| b.is_ascii_hexdigit()))
+        || (hash.len() == 32
+            && hash
+                .bytes()
+                .all(|b| matches!(b, b'2'..=b'7' | b'a'..=b'z' | b'A'..=b'Z')))
+}
+
+pub(crate) fn build_magnet(info_hash: &str, name: &str) -> String {
+    let mut magnet = format!("magnet:?xt=urn:btih:{info_hash}&dn={}", urlencode(name));
+    for t in MAGNET_TRACKERS {
+        magnet.push_str("&tr=");
+        magnet.push_str(&urlencode(t));
+    }
+    magnet
+}
+
 pub(crate) fn urlencode(s: &str) -> String {
     s.chars()
         .map(|c| {

breadarrd/src/sources/torrents_csv.rs

@@ -0,0 +1,130 @@
+use anyhow::{Context, Result};
+use async_trait::async_trait;
+use serde::Deserialize;
+
+use super::{build_magnet, is_valid_info_hash, urlencode, RawReleaseItem, ReleaseSource};
+
+/// Public JSON search over the torrents.csv DHT dump. Same grab shape as
+/// TPB (info-hash → magnet, no HTML): used as the first general-content
+/// fallback when apibay is down or returns nothing. Seeders are scrape
+/// snapshots, not live tracker data, so a high number can still stall —
+/// the existing seeder gate still applies.
+pub struct TorrentsCsvSource {
+    api_url: String,
+    client: reqwest::Client,
+}
+
+#[derive(Deserialize)]
+struct CsvResponse {
+    #[serde(default)]
+    torrents: Vec<CsvTorrent>,
+}
+
+#[derive(Deserialize)]
+struct CsvTorrent {
+    infohash: String,
+    name: String,
+    size_bytes: Option<u64>,
+    seeders: Option<i64>,
+    leechers: Option<i64>,
+}
+
+impl TorrentsCsvSource {
+    pub fn new(api_url: impl Into<String>) -> Self {
+        Self {
+            api_url: api_url.into(),
+            client: reqwest::Client::builder()
+                .timeout(std::time::Duration::from_secs(30))
+                .build()
+                .expect("reqwest client build"),
+        }
+    }
+}
+
+fn as_u32(n: Option<i64>) -> Option<u32> {
+    n.and_then(|v| u32::try_from(v.max(0)).ok())
+}
+
+#[async_trait]
+impl ReleaseSource for TorrentsCsvSource {
+    async fn fetch(&self, query: Option<&str>) -> Result<Vec<RawReleaseItem>> {
+        let Some(query) = query else {
+            anyhow::bail!(
+                "TorrentsCsvSource requires a search query (this is a search-driven source, not a feed)"
+            );
+        };
+        let sep = if self.api_url.contains('?') { '&' } else { '?' };
+        let url = format!(
+            "{}{sep}q={}&size=25",
+            self.api_url.trim_end_matches('/'),
+            urlencode(query)
+        );
+        let parsed: CsvResponse = self
+            .client
+            .get(&url)
+            .send()
+            .await
+            .with_context(|| format!("request to {url} failed"))?
+            .error_for_status()
+            .with_context(|| format!("{url} returned an error status"))?
+            .json()
+            .await
+            .context("failed to parse torrents-csv response as JSON")?;
+
+        Ok(parsed
+            .torrents
+            .into_iter()
+            .filter(|t| is_valid_info_hash(&t.infohash))
+            .map(|t| RawReleaseItem {
+                title: t.name.clone(),
+                link: build_magnet(&t.infohash, &t.name),
+                guid: t.infohash,
+                size_bytes: t.size_bytes,
+                seeders: as_u32(t.seeders),
+                leechers: as_u32(t.leechers),
+            })
+            .collect())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn parses_a_real_captured_response() {
+        let body = r#"{
+            "torrents": [
+                {
+                    "infohash": "ed0da850c273e3e15a819bdcbbf418bc85107ec8",
+                    "name": "Dune (2021) [1080p] [WEBRip]",
+                    "size_bytes": 2947989023,
+                    "seeders": 746,
+                    "leechers": 38
+                },
+                {
+                    "infohash": "not-a-hash",
+                    "name": "garbage",
+                    "size_bytes": 1,
+                    "seeders": 0,
+                    "leechers": 0
+                }
+            ]
+        }"#;
+        let parsed: CsvResponse = serde_json::from_str(body).unwrap();
+        let items: Vec<_> = parsed
+            .torrents
+            .into_iter()
+            .filter(|t| is_valid_info_hash(&t.infohash))
+            .collect();
+        assert_eq!(items.len(), 1);
+        assert_eq!(items[0].name, "Dune (2021) [1080p] [WEBRip]");
+        assert_eq!(items[0].seeders, Some(746));
+    }
+
+    #[test]
+    fn empty_payload_is_no_results_not_an_error() {
+        let parsed: CsvResponse = serde_json::from_str(r#"{"torrents":[]}"#).unwrap();
+        assert!(parsed.torrents.is_empty());
+    }
+}

breadarrd/src/sources/tpb.rs

@@ -2,18 +2,7 @@ use anyhow::{Context, Result};
 use async_trait::async_trait;
 use serde::Deserialize;
 
-use super::{urlencode, RawReleaseItem, ReleaseSource};
+use super::{build_magnet, is_valid_info_hash, urlencode, RawReleaseItem, ReleaseSource};
-
-/// A valid BitTorrent v1 info_hash: 40 hex chars or 32 base32 chars — same
-/// shape `qbit::extract_btih` accepts out of a magnet URI. Checked before
-/// building a magnet from `info_hash` at all: apibay is normally reliable,
-/// but a malformed value would otherwise silently produce a magnet
-/// `extract_btih` can't parse back out, downgrading that grab to the slow
-/// ~30s `torrents/info` polling path with no visible error anywhere.
-fn is_valid_info_hash(hash: &str) -> bool {
-    (hash.len() == 40 && hash.bytes().all(|b| b.is_ascii_hexdigit()))
-        || (hash.len() == 32 && hash.bytes().all(|b| matches!(b, b'2'..=b'7' | b'a'..=b'z' | b'A'..=b'Z')))
-}
 
 /// A community-run JSON API mirror of The Pirate Bay's search — unlike
 /// 1337x, this is a genuine machine-readable API (not HTML scraping), and
@@ -38,23 +27,6 @@ struct TpbResult {
     size: String,
 }
 
-const TRACKERS: &[&str] = &[
-    "udp://tracker.opentrackr.org:1337/announce",
-    "udp://open.stealth.si:80/announce",
-    "udp://tracker.torrent.eu.org:451/announce",
-    "udp://tracker.openbittorrent.com:6969/announce",
-    "udp://exodus.desync.com:6969/announce",
-];
-
-fn build_magnet(info_hash: &str, name: &str) -> String {
-    let mut magnet = format!("magnet:?xt=urn:btih:{info_hash}&dn={}", urlencode(name));
-    for t in TRACKERS {
-        magnet.push_str("&tr=");
-        magnet.push_str(&urlencode(t));
-    }
-    magnet
-}
-
 impl TpbSource {
     pub fn new(api_url: impl Into<String>) -> Self {
         Self {
@@ -154,7 +126,9 @@ mod tests {
 
     #[test]
     fn is_valid_info_hash_accepts_both_real_shapes() {
-        assert!(is_valid_info_hash("8F87C7C186172F17E35F4512BB1A3E93B614ADED")); // 40 hex
+        assert!(is_valid_info_hash(
+            "8F87C7C186172F17E35F4512BB1A3E93B614ADED"
+        )); // 40 hex
         assert!(is_valid_info_hash("abcdefghijklmnopqrstuvwxyz234567")); // 32 base32
     }
 
@@ -167,7 +141,11 @@ mod tests {
     fn is_valid_info_hash_rejects_malformed_values() {
         assert!(!is_valid_info_hash(""));
         assert!(!is_valid_info_hash("too-short"));
-        assert!(!is_valid_info_hash("not-a-hex-string-at-all-nope!!!!!!!!!!!!")); // 40 chars, non-hex
+        assert!(!is_valid_info_hash(
-        assert!(!is_valid_info_hash("8F87C7C186172F17E35F4512BB1A3E93B614ADE")); // 39 hex chars
+            "not-a-hex-string-at-all-nope!!!!!!!!!!!!"
+        )); // 40 chars, non-hex
+        assert!(!is_valid_info_hash(
+            "8F87C7C186172F17E35F4512BB1A3E93B614ADE"
+        )); // 39 hex chars
     }
 }

breadarrd/src/sources/yts.rs

@@ -0,0 +1,194 @@
+use anyhow::{Context, Result};
+use async_trait::async_trait;
+use serde::Deserialize;
+
+use super::{build_magnet, is_valid_info_hash, urlencode, RawReleaseItem, ReleaseSource};
+
+/// YTS movie API. `yts.mx` itself no longer resolves (checked 2026-07-12
+/// and again 2026-08-16); the `yts.lt` / `yts.am` hosts still serve the
+/// v2 JSON API, which is why the default URL is a working mirror rather
+/// than the brand domain. Movies only — each hit expands into one
+/// `RawReleaseItem` per quality so the scorer sees 720p/1080p/2160p as
+/// distinct candidates, same as if they were separate TPB rows.
+pub struct YtsSource {
+    api_url: String,
+    client: reqwest::Client,
+}
+
+#[derive(Deserialize)]
+struct YtsResponse {
+    status: String,
+    data: Option<YtsData>,
+}
+
+#[derive(Deserialize, Default)]
+struct YtsData {
+    #[serde(default)]
+    movies: Vec<YtsMovie>,
+}
+
+#[derive(Deserialize)]
+struct YtsMovie {
+    title: String,
+    year: Option<i64>,
+    #[serde(default)]
+    torrents: Vec<YtsTorrent>,
+}
+
+#[derive(Deserialize)]
+struct YtsTorrent {
+    hash: String,
+    quality: Option<String>,
+    #[serde(rename = "type")]
+    source_type: Option<String>,
+    video_codec: Option<String>,
+    seeds: Option<i64>,
+    peers: Option<i64>,
+    size_bytes: Option<u64>,
+}
+
+impl YtsSource {
+    pub fn new(api_url: impl Into<String>) -> Self {
+        Self {
+            api_url: api_url.into(),
+            client: reqwest::Client::builder()
+                .timeout(std::time::Duration::from_secs(30))
+                .build()
+                .expect("reqwest client build"),
+        }
+    }
+}
+
+fn as_u32(n: Option<i64>) -> Option<u32> {
+    n.and_then(|v| u32::try_from(v.max(0)).ok())
+}
+
+/// Builds a release title the existing parser can read quality/source/codec
+/// out of — YTS stores those as structured fields, not in `title`.
+fn release_title(movie: &YtsMovie, torrent: &YtsTorrent) -> String {
+    let mut title = movie.title.clone();
+    if let Some(year) = movie.year {
+        title.push_str(&format!(" ({year})"));
+    }
+    for part in [
+        torrent.quality.as_deref(),
+        torrent.source_type.as_deref(),
+        torrent.video_codec.as_deref(),
+    ]
+    .into_iter()
+    .flatten()
+    {
+        if !part.is_empty() {
+            title.push_str(&format!(" [{part}]"));
+        }
+    }
+    title
+}
+
+#[async_trait]
+impl ReleaseSource for YtsSource {
+    async fn fetch(&self, query: Option<&str>) -> Result<Vec<RawReleaseItem>> {
+        let Some(query) = query else {
+            anyhow::bail!(
+                "YtsSource requires a search query (this is a search-driven source, not a feed)"
+            );
+        };
+        let sep = if self.api_url.contains('?') { '&' } else { '?' };
+        let url = format!(
+            "{}{sep}query_term={}&limit=20&sort_by=seeds",
+            self.api_url.trim_end_matches('/'),
+            urlencode(query)
+        );
+        let parsed: YtsResponse = self
+            .client
+            .get(&url)
+            .send()
+            .await
+            .with_context(|| format!("request to {url} failed"))?
+            .error_for_status()
+            .with_context(|| format!("{url} returned an error status"))?
+            .json()
+            .await
+            .context("failed to parse YTS response as JSON")?;
+        anyhow::ensure!(
+            parsed.status == "ok",
+            "YTS returned status {:?}",
+            parsed.status
+        );
+
+        let movies = parsed.data.unwrap_or_default().movies;
+        let mut items = Vec::new();
+        for movie in movies {
+            for torrent in &movie.torrents {
+                if !is_valid_info_hash(&torrent.hash) {
+                    continue;
+                }
+                let title = release_title(&movie, torrent);
+                items.push(RawReleaseItem {
+                    title: title.clone(),
+                    link: build_magnet(&torrent.hash, &title),
+                    guid: torrent.hash.clone(),
+                    size_bytes: torrent.size_bytes,
+                    seeders: as_u32(torrent.seeds),
+                    leechers: as_u32(torrent.peers),
+                });
+            }
+        }
+        Ok(items)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    const SAMPLE: &str = r#"{
+        "status": "ok",
+        "data": {
+            "movie_count": 1,
+            "movies": [{
+                "title": "Dune: Part One",
+                "year": 2021,
+                "torrents": [
+                    {
+                        "hash": "DEB6929BEEB09ADCBD14DC4D6081F7E6B297B88C",
+                        "quality": "1080p",
+                        "type": "web",
+                        "video_codec": "x264",
+                        "seeds": 12,
+                        "peers": 3,
+                        "size_bytes": 2147483648
+                    },
+                    {
+                        "hash": "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",
+                        "quality": "720p",
+                        "type": "bluray",
+                        "video_codec": "x265",
+                        "seeds": 4,
+                        "peers": 1,
+                        "size_bytes": 1073741824
+                    }
+                ]
+            }]
+        }
+    }"#;
+
+    #[test]
+    fn expands_one_movie_into_per_quality_rows() {
+        let parsed: YtsResponse = serde_json::from_str(SAMPLE).unwrap();
+        assert_eq!(parsed.status, "ok");
+        let movie = &parsed.data.unwrap().movies[0];
+        assert_eq!(movie.torrents.len(), 2);
+        let t0 = release_title(movie, &movie.torrents[0]);
+        assert_eq!(t0, "Dune: Part One (2021) [1080p] [web] [x264]");
+        let t1 = release_title(movie, &movie.torrents[1]);
+        assert_eq!(t1, "Dune: Part One (2021) [720p] [bluray] [x265]");
+    }
+
+    #[test]
+    fn missing_movies_array_is_empty_not_an_error() {
+        let parsed: YtsResponse =
+            serde_json::from_str(r#"{"status":"ok","data":{"movie_count":0}}"#).unwrap();
+        assert!(parsed.data.unwrap_or_default().movies.is_empty());
+    }
+}

config.example.toml

@@ -72,6 +72,12 @@ import_poll_interval_secs = 60
 # Community JSON API mirror of The Pirate Bay — primary general-content
 # search source (movies + non-anime TV).
 tpb_api_url = "https://apibay.org/q.php"
+# torrents.csv DHT-dump search — first fallback when TPB fails or returns
+# nothing. Same hash-to-magnet grab; movies and TV.
+torrents_csv_url = "https://torrents-csv.com/service/search"
+# YTS v2 list_movies JSON — movie-only fallback. yts.mx does not resolve;
+# yts.lt is a working host as of 2026-08-16.
+yts_api_url = "https://yts.lt/api/v2/list_movies.json"
 # Search-driven acquisition (movies + non-anime TV via 1337x, anime movies
 # via nyaa's search mode) — unlike the nyaa RSS feed watch above, this
 # actively queries a Cloudflare-fronted service with a ban history, so keep
@@ -96,6 +102,7 @@ upgrade_min_score_gain = 5.0
 # tried in a fixed fallback order), with a failing mirror demoted into a
 # cooldown rather than re-probed on the very next search.
 torrent_1337x_mirrors = [
+  "https://www.1337xx.to",
   "https://13377x.info",
   "https://13377x.email",
   "https://1337xto.info",