feat: Add auto-trigger for LLM analysis

Introduces a new module `auto_trigger` responsible for opportunistically
initiating LLM analysis jobs. This mechanism scans case directories for
new or updated recordings and transcripts, automatically creating and
queuing analysis jobs when appropriate.

Key components:
- `evaluate_case`: Pure function to determine if a case is ready for
  analysis based on recording status, transcriptions, document
  modification times, and existing job/failure markers.
- `try_enqueue`: Orchestrates the analysis job creation and queuing
  process if `evaluate_case` returns `AutoDecision::Enqueue`.
- `try_enqueue_all_for_user`: Iterates through all user cases to trigger
  `try_enqueue` for eligible ones.
- `write_failure_marker`/`remove_failure_marker`: Handles persistent
  recording of analysis failures and their cleanup.

This feature aims to reduce manual intervention by automatically
analyzing new or changed case data.

server/src/routes/mod.rs

@@ -1,5 +1,5 @@
 mod bulk;
-mod case_actions;
+pub(crate) mod case_actions;
 mod debug;
 mod events;
 mod health;

server/src/routes/user_web.rs

@@ -10,11 +10,14 @@ use tracing::{info, warn};
 use time::{Date, OffsetDateTime, format_description::well_known::Rfc3339};
 
 use crate::PipelineState;
-use crate::analyze::{ANALYSIS_INPUT_FILE, DOCUMENT_FILE, recovery as analyze_recovery};
+use crate::analyze::{
+    ANALYSIS_INPUT_FILE, DOCUMENT_FILE, auto_trigger, recovery as analyze_recovery,
+};
 use crate::auth::AuthenticatedWebUser;
 use crate::case_id::{CaseId, CaseIdPath};
 use crate::config::Config;
 use crate::error::AppError;
+use crate::events::EventSender;
 use crate::routes::case_actions::read_document;
 use crate::routes::web::{RecordingView, scan_recordings};
 use crate::transcribe::recovery as transcribe_recovery;
@@ -228,9 +231,16 @@ pub async fn handle_my_cases(
     user: AuthenticatedWebUser,
     State(config): State<Arc<Config>>,
     State(pipeline): State<PipelineState>,
+    State(events_tx): State<EventSender>,
 ) -> Result<Html<String>, AppError> {
     let user_root = config.data_path.join(&user.slug);
     pipeline.heal_orphans_if_idle(&user_root, &user.slug).await;
+    // Auto-analysis: hand every eligible case to the worker before we
+    // render. The common case is "nothing to do" and costs a handful of
+    // stat-calls per case; actual enqueues happen only when pre-conditions
+    // flip (new transcripts in, stale document, ...).
+    auto_trigger::try_enqueue_all_for_user(&user_root, &pipeline.analyze_tx, &config, &events_tx)
+        .await;
 
     let busy = pipeline.analyze_busy.0.load(Ordering::Acquire);
     let cases = scan_user_cases(&config, &user.slug, busy).await;
@@ -263,6 +273,7 @@ pub async fn handle_case_page(
     user: AuthenticatedWebUser,
     State(config): State<Arc<Config>>,
     State(pipeline): State<PipelineState>,
+    State(events_tx): State<EventSender>,
     CaseIdPath(case_id): CaseIdPath,
 ) -> Result<Html<String>, AppError> {
     let user_root = config.data_path.join(&user.slug);
@@ -275,6 +286,10 @@ pub async fn handle_case_page(
         "case page (possible IDOR probe)",
     )
     .await?;
+    // Auto-analysis trigger for deep-link navigation: same evaluation as
+    // the list view. Document render below still wins the race if the
+    // worker happens to finish synchronously.
+    auto_trigger::try_enqueue(&case_dir, &pipeline.analyze_tx, &config, &events_tx).await;
     let case_id_str = case_id.to_string();
     info!(slug = %user.slug, case_id = %case_id, "case page viewed");