Merge pull request #1182 from nearai/staging-promote/579c4fdb-23095333790

chore: promote staging to staging-promote/17706632-23094430993 (2026-03-14 20:03 UTC)

.gitignore

@@ -14,6 +14,10 @@
 
 target/
 
+# Python
+__pycache__/
+*.pyc
+
 # Benchmark results (local runs, not committed)
 bench-results/
 

Cargo.lock

@@ -151,7 +151,7 @@ version = "1.1.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "40c48f72fd53cd289104fc64099abca73db4166ad86ea0b4341abe65af83dadc"
 dependencies = [
- "windows-sys 0.61.2",
+ "windows-sys 0.60.2",
 ]
 
 [[package]]
@@ -162,7 +162,7 @@ checksum = "291e6a250ff86cd4a820112fb8898808a366d8f9f58ce16d1f538353ad55747d"
 dependencies = [
  "anstyle",
  "once_cell_polyfill",
- "windows-sys 0.61.2",
+ "windows-sys 0.60.2",
 ]
 
 [[package]]
@@ -2077,7 +2077,7 @@ dependencies = [
  "libc",
  "option-ext",
  "redox_users 0.5.2",
- "windows-sys 0.61.2",
+ "windows-sys 0.59.0",
 ]
 
 [[package]]
@@ -2264,7 +2264,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
 dependencies = [
  "libc",
- "windows-sys 0.61.2",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -4089,7 +4089,7 @@ version = "0.50.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7957b9740744892f114936ab4a57b3f487491bbeafaf8083688b16841a4240e5"
 dependencies = [
- "windows-sys 0.61.2",
+ "windows-sys 0.59.0",
 ]
 
 [[package]]
@@ -5433,7 +5433,7 @@ dependencies = [
  "errno",
  "libc",
  "linux-raw-sys 0.12.1",
- "windows-sys 0.61.2",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -6115,7 +6115,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3a766e1110788c36f4fa1c2b71b387a7815aa65f88ce0229841826633d93723e"
 dependencies = [
  "libc",
- "windows-sys 0.61.2",
+ "windows-sys 0.60.2",
 ]
 
 [[package]]
@@ -6337,10 +6337,10 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "32497e9a4c7b38532efcdebeef879707aa9f794296a4f0244f6f69e9bc8574bd"
 dependencies = [
  "fastrand",
- "getrandom 0.4.2",
+ "getrandom 0.3.4",
  "once_cell",
  "rustix 1.1.4",
- "windows-sys 0.61.2",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -7134,13 +7134,13 @@ checksum = "2896d95c02a80c6d6a5d6e953d479f5ddf2dfdb6a244441010e373ac0fb88971"
 
 [[package]]
 name = "uds_windows"
-version = "1.2.0"
+version = "1.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "51b70b87d15e91f553711b40df3048faf27a7a04e01e0ddc0cf9309f0af7c2ca"
+checksum = "f2f6fb2847f6742cd76af783a2a2c49e9375d0a111c7bef6f71cd9e738c72d6e"
 dependencies = [
  "memoffset",
  "tempfile",
- "windows-sys 0.61.2",
+ "windows-sys 0.60.2",
 ]
 
 [[package]]
@@ -7996,7 +7996,7 @@ version = "0.1.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c2a7b1c03c876122aa43f3020e6c3c3ee5c05081c9a00739faf7503aeba10d22"
 dependencies = [
- "windows-sys 0.61.2",
+ "windows-sys 0.48.0",
 ]
 
 [[package]]

registry/channels/discord.json

@@ -2,7 +2,7 @@
   "name": "discord",
   "display_name": "Discord Channel",
   "kind": "channel",
-  "version": "0.2.1",
+  "version": "0.2.0",
   "wit_version": "0.3.0",
   "description": "Talk to your agent in Discord",
   "keywords": [

registry/tools/github.json

@@ -2,7 +2,7 @@
   "name": "github",
   "display_name": "GitHub",
   "kind": "tool",
-  "version": "0.2.1",
+  "version": "0.2.0",
   "wit_version": "0.3.0",
   "description": "GitHub integration for issues, PRs, repos, and code search",
   "keywords": [

registry/tools/web-search.json

@@ -2,7 +2,7 @@
   "name": "web-search",
   "display_name": "Web Search",
   "kind": "tool",
-  "version": "0.2.1",
+  "version": "0.2.0",
   "wit_version": "0.3.0",
   "description": "Search the web using Brave Search API",
   "keywords": [

src/agent/routine.rs

@@ -538,11 +538,174 @@ pub fn next_cron_fire(
     }
 }
 
+/// Describe common routine cron patterns in plain English.
+///
+/// Falls back to `cron: <raw>` for malformed or complex expressions.
+pub fn describe_cron(schedule: &str, timezone: Option<&str>) -> String {
+    fn fallback(raw: &str) -> String {
+        if raw.trim().is_empty() {
+            "cron: (empty)".to_string()
+        } else {
+            format!("cron: {}", raw.trim())
+        }
+    }
+
+    fn parse_u8_token(token: &str) -> Option<u8> {
+        token.parse::<u8>().ok()
+    }
+
+    fn parse_step(token: &str) -> Option<u8> {
+        token
+            .strip_prefix("*/")
+            .and_then(parse_u8_token)
+            .filter(|n| *n > 0)
+    }
+
+    fn weekday_name(dow: &str) -> Option<&'static str> {
+        let normalized = dow.trim().to_ascii_uppercase();
+        match normalized.as_str() {
+            "MON" | "1" => Some("Monday"),
+            "TUE" | "2" => Some("Tuesday"),
+            "WED" | "3" => Some("Wednesday"),
+            "THU" | "4" => Some("Thursday"),
+            "FRI" | "5" => Some("Friday"),
+            "SAT" | "6" => Some("Saturday"),
+            "SUN" | "0" | "7" => Some("Sunday"),
+            _ => None,
+        }
+    }
+
+    fn format_time(hour: u8, minute: u8) -> String {
+        if hour == 0 && minute == 0 {
+            return "midnight".to_string();
+        }
+        let (display_hour, am_pm) = match hour {
+            0 => (12, "AM"),
+            1..=11 => (hour, "AM"),
+            12 => (12, "PM"),
+            _ => (hour - 12, "PM"),
+        };
+        format!("{display_hour}:{minute:02} {am_pm}")
+    }
+
+    fn ordinal(n: u8) -> String {
+        let suffix = if (11..=13).contains(&(n % 100)) {
+            "th"
+        } else {
+            match n % 10 {
+                1 => "st",
+                2 => "nd",
+                3 => "rd",
+                _ => "th",
+            }
+        };
+        format!("{n}{suffix}")
+    }
+
+    fn describe_inner(raw: &str) -> Option<String> {
+        let fields: Vec<&str> = raw.split_whitespace().collect();
+        let (sec, min, hour, dom, month, dow, year) = match fields.len() {
+            5 => (
+                "0", fields[0], fields[1], fields[2], fields[3], fields[4], None,
+            ),
+            6 => (
+                fields[0], fields[1], fields[2], fields[3], fields[4], fields[5], None,
+            ),
+            7 => (
+                fields[0],
+                fields[1],
+                fields[2],
+                fields[3],
+                fields[4],
+                fields[5],
+                Some(fields[6]),
+            ),
+            _ => return None,
+        };
+
+        if year.is_some_and(|v| v != "*") {
+            return None;
+        }
+
+        if sec == "0"
+            && hour == "*"
+            && dom == "*"
+            && month == "*"
+            && dow == "*"
+            && let Some(step) = parse_step(min)
+        {
+            return Some(match step {
+                1 => "Every minute".to_string(),
+                n => format!("Every {n} minutes"),
+            });
+        }
+
+        if sec == "0"
+            && min == "0"
+            && dom == "*"
+            && month == "*"
+            && dow == "*"
+            && let Some(step) = parse_step(hour)
+        {
+            return Some(match step {
+                1 => "Every hour".to_string(),
+                n => format!("Every {n} hours"),
+            });
+        }
+
+        let hour = parse_u8_token(hour).filter(|h| *h <= 23)?;
+        let minute = parse_u8_token(min).filter(|m| *m <= 59)?;
+        let time = format_time(hour, minute);
+        let time_phrase = if time == "midnight" {
+            "at midnight".to_string()
+        } else {
+            format!("at {time}")
+        };
+
+        if sec == "0" && dom == "*" && month == "*" && dow == "*" {
+            return Some(format!("Daily {time_phrase}"));
+        }
+
+        if sec == "0" && dom == "*" && month == "*" && dow.eq_ignore_ascii_case("MON-FRI") {
+            return Some(format!("Weekdays {time_phrase}"));
+        }
+
+        if sec == "0"
+            && dom == "*"
+            && month == "*"
+            && let Some(day_name) = weekday_name(dow)
+        {
+            return Some(format!("Every {day_name} {time_phrase}"));
+        }
+
+        if sec == "0"
+            && month == "*"
+            && dow == "*"
+            && let Some(day_of_month) = parse_u8_token(dom).filter(|d| (1..=31).contains(d))
+        {
+            return Some(format!(
+                "{} of every month {time_phrase}",
+                ordinal(day_of_month)
+            ));
+        }
+
+        None
+    }
+
+    let mut description = describe_inner(schedule).unwrap_or_else(|| fallback(schedule));
+    if let Some(tz) = timezone.map(str::trim).filter(|tz| !tz.is_empty()) {
+        description.push_str(" (");
+        description.push_str(tz);
+        description.push(')');
+    }
+    description
+}
+
 #[cfg(test)]
 mod tests {
     use crate::agent::routine::{
         MAX_TOOL_ROUNDS_LIMIT, RoutineAction, RoutineGuardrails, RunStatus, Trigger, content_hash,
-        next_cron_fire,
+        describe_cron, next_cron_fire,
     };
 
     #[test]
@@ -698,6 +861,40 @@ mod tests {
         assert_ne!(next_utc, next_est, "timezone should shift the fire time");
     }
 
+    #[test]
+    fn test_describe_cron_common_patterns() {
+        let cases = vec![
+            ("0 */30 * * * *", None, "Every 30 minutes"),
+            ("0 0 9 * * *", None, "Daily at 9:00 AM"),
+            ("0 0 9 * * MON-FRI", None, "Weekdays at 9:00 AM"),
+            ("0 0 */2 * * *", None, "Every 2 hours"),
+            ("0 0 0 * * *", None, "Daily at midnight"),
+            ("0 0 9 * * 1", None, "Every Monday at 9:00 AM"),
+            ("0 0 9 1 * *", None, "1st of every month at 9:00 AM"),
+            (
+                "0 0 9 * * MON-FRI",
+                Some("America/New_York"),
+                "Weekdays at 9:00 AM (America/New_York)",
+            ),
+            ("1 2 3 4 5 6", None, "cron: 1 2 3 4 5 6"),
+        ];
+
+        for (schedule, timezone, expected) in cases {
+            let actual = describe_cron(schedule, timezone);
+            assert_eq!(actual, expected); // safety: test-only assertion in #[cfg(test)] module
+        }
+    }
+
+    #[test]
+    fn test_describe_cron_edge_cases() {
+        assert_eq!(describe_cron("", None), "cron: (empty)"); // safety: test-only assertion in #[cfg(test)] module
+        assert_eq!(describe_cron("not a cron", None), "cron: not a cron"); // safety: test-only assertion in #[cfg(test)] module
+        let weekdays_5_field = describe_cron("0 9 * * MON-FRI", None);
+        assert_eq!(weekdays_5_field, "Weekdays at 9:00 AM"); // safety: test-only assertion in #[cfg(test)] module
+        let weekdays_7_field = describe_cron("0 0 9 * * MON-FRI *", None);
+        assert_eq!(weekdays_7_field, "Weekdays at 9:00 AM"); // safety: test-only assertion in #[cfg(test)] module
+    }
+
     #[test]
     fn test_guardrails_default() {
         let g = RoutineGuardrails::default();

src/agent/routine_engine.rs

@@ -139,6 +139,32 @@ impl RoutineEngine {
         let cache = self.event_cache.read().await;
         let mut fired = 0;
 
+        // Collect routine IDs for batch query
+        let routine_ids: Vec<Uuid> = cache
+            .iter()
+            .filter_map(|matcher| match matcher {
+                EventMatcher::Message { routine, .. } => Some(routine.id),
+                EventMatcher::System { .. } => None,
+            })
+            .collect();
+
+        if routine_ids.is_empty() {
+            return 0;
+        }
+
+        // Single batch query instead of N queries
+        let concurrent_counts = match self
+            .store
+            .count_running_routine_runs_batch(&routine_ids)
+            .await
+        {
+            Ok(counts) => counts,
+            Err(e) => {
+                tracing::error!("Failed to batch-load concurrent counts: {}", e);
+                return 0;
+            }
+        };
+
         for matcher in cache.iter() {
             let (routine, re) = match matcher {
                 EventMatcher::Message { routine, regex } => (routine, regex),
@@ -164,8 +190,9 @@ impl RoutineEngine {
                 continue;
             }
 
-            // Concurrent run check
-            if !self.check_concurrent(routine).await {
+            // Concurrent run check (using batch-loaded counts)
+            let running_count = concurrent_counts.get(&routine.id).copied().unwrap_or(0);
+            if running_count >= routine.guardrails.max_concurrent as i64 {
                 tracing::trace!(routine = %routine.name, "Skipped: max concurrent reached");
                 continue;
             }
@@ -197,6 +224,35 @@ impl RoutineEngine {
         let cache = self.event_cache.read().await;
         let mut fired = 0;
 
+        // Collect routine IDs for batch query
+        let routine_ids: Vec<Uuid> = cache
+            .iter()
+            .filter_map(|matcher| match matcher {
+                EventMatcher::System { routine } => Some(routine.id),
+                EventMatcher::Message { .. } => None,
+            })
+            .collect();
+
+        if routine_ids.is_empty() {
+            return 0;
+        }
+
+        // Single batch query instead of N queries
+        let concurrent_counts = match self
+            .store
+            .count_running_routine_runs_batch(&routine_ids)
+            .await
+        {
+            Ok(counts) => counts,
+            Err(e) => {
+                tracing::error!(
+                    "Failed to batch-load concurrent counts for system events: {}",
+                    e
+                );
+                return 0;
+            }
+        };
+
         for matcher in cache.iter() {
             let routine = match matcher {
                 EventMatcher::System { routine } => routine,
@@ -248,7 +304,9 @@ impl RoutineEngine {
                 continue;
             }
 
-            if !self.check_concurrent(routine).await {
+            // Concurrent run check (using batch-loaded counts)
+            let running_count = concurrent_counts.get(&routine.id).copied().unwrap_or(0);
+            if running_count >= routine.guardrails.max_concurrent as i64 {
                 tracing::debug!(routine = %routine.name, "Skipped: max concurrent reached");
                 continue;
             }
@@ -925,7 +983,8 @@ async fn execute_lightweight_with_tools(
                 .tool_definitions_excluding(ROUTINE_TOOL_DENYLIST)
                 .await;
 
-            let request = ToolCompletionRequest::new(messages.clone(), tool_defs)
+            let request_messages = snapshot_messages_for_tool_iteration(&messages);
+            let request = ToolCompletionRequest::new(request_messages, tool_defs)
                 .with_max_tokens(effective_max_tokens)
                 .with_temperature(0.3);
 
@@ -1001,6 +1060,31 @@ async fn execute_lightweight_with_tools(
     }
 }
 
+// Bound per-iteration context copy cost for lightweight tool loops.
+const MAX_TOOL_LOOP_MESSAGES: usize = 32;
+
+fn snapshot_messages_for_tool_iteration(messages: &[ChatMessage]) -> Vec<ChatMessage> {
+    if messages.len() <= MAX_TOOL_LOOP_MESSAGES {
+        return messages.to_vec();
+    }
+
+    let mut snapshot = Vec::with_capacity(MAX_TOOL_LOOP_MESSAGES);
+
+    if let Some(first) = messages.first()
+        && first.role == crate::llm::Role::System
+    {
+        snapshot.push(first.clone());
+        let tail_len = MAX_TOOL_LOOP_MESSAGES - 1;
+        let tail_start = (messages.len() - tail_len).max(1);
+        snapshot.extend_from_slice(&messages[tail_start..]);
+    } else {
+        let tail_start = messages.len() - MAX_TOOL_LOOP_MESSAGES;
+        snapshot.extend_from_slice(&messages[tail_start..]);
+    }
+
+    snapshot
+}
+
 /// Tools that must never be callable from lightweight routines.
 ///
 /// These tools pose autonomy-escalation risks: a routine could self-replicate,
@@ -1386,4 +1470,33 @@ mod tests {
         let out = super::truncate(input, 5);
         assert_eq!(out, "abcde...");
     }
+
+    #[test]
+    fn test_snapshot_messages_keeps_system_and_recent_tail() {
+        let mut messages = vec![crate::llm::ChatMessage::system("sys")];
+        for i in 0..80 {
+            messages.push(crate::llm::ChatMessage::user(format!("u{i}")));
+        }
+
+        let snapshot = super::snapshot_messages_for_tool_iteration(&messages);
+        assert_eq!(snapshot.len(), super::MAX_TOOL_LOOP_MESSAGES); // safety: test-only no-panics CI false positive
+        assert_eq!(snapshot[0].role, crate::llm::Role::System); // safety: test-only no-panics CI false positive
+        assert_eq!(snapshot[0].content, "sys"); // safety: test-only no-panics CI false positive
+        let last_content = snapshot.last().map(|m| m.content.as_str());
+        assert_eq!(last_content, Some("u79")); // safety: test-only no-panics CI false positive
+    }
+
+    #[test]
+    fn test_snapshot_messages_unchanged_when_within_limit() {
+        let messages = vec![
+            crate::llm::ChatMessage::system("sys"),
+            crate::llm::ChatMessage::user("a"),
+            crate::llm::ChatMessage::assistant("b"),
+        ];
+        let snapshot = super::snapshot_messages_for_tool_iteration(&messages);
+        assert_eq!(snapshot.len(), messages.len()); // safety: test-only no-panics CI false positive
+        assert_eq!(snapshot[0].role, crate::llm::Role::System); // safety: test-only no-panics CI false positive
+        assert_eq!(snapshot[1].content, "a"); // safety: test-only no-panics CI false positive
+        assert_eq!(snapshot[2].content, "b"); // safety: test-only no-panics CI false positive
+    }
 }

src/agent/scheduler.rs

@@ -179,27 +179,33 @@ impl Scheduler {
             })
             .unwrap_or(self.config.max_tokens_per_job);
 
-        // Apply both metadata and token budget in one closure (Issue #813: atomic update)
-        if let Some(meta) = metadata {
+        // Apply both metadata and token budget in one closure (Issue #813: atomic update).
+        // Use update_context_and_get to ensure atomicity: no gap where concurrent workers
+        // can modify the context between update and DB persist (Issue #807).
+        let ctx = if let Some(meta) = metadata {
             self.context_manager
-                .update_context(job_id, |ctx| {
+                .update_context_and_get(job_id, |ctx| {
                     ctx.metadata = meta;
                     if max_tokens > 0 {
                         ctx.max_tokens = max_tokens;
                     }
                 })
-                .await?;
+                .await?
         } else if max_tokens > 0 {
             self.context_manager
-                .update_context(job_id, |ctx| {
+                .update_context_and_get(job_id, |ctx| {
                     ctx.max_tokens = max_tokens;
                 })
-                .await?;
-        }
+                .await?
+        } else {
+            // No metadata or token budget to set; get the initial context
+            self.context_manager.get_context(job_id).await?
+        };
 
-        // Persist to DB before scheduling so the worker's FK references are valid
+        // Persist to DB before scheduling so the worker's FK references are valid.
+        // The context was read under the same lock as the update (atomic), preventing
+        // concurrent worker interference (Issue #807: non-transactional context updates).
         if let Some(ref store) = self.store {
-            let ctx = self.context_manager.get_context(job_id).await?;
             store.save_job(&ctx).await.map_err(|e| JobError::Failed {
                 id: job_id,
                 reason: format!("failed to persist job: {e}"),
@@ -832,6 +838,24 @@ mod tests {
         );
     }
 
+    #[tokio::test]
+    async fn test_dispatch_job_no_metadata_no_user_tokens_edge_case() {
+        // Edge case coverage: when metadata=None AND max_tokens=0 (config),
+        // the else branch calls get_context() directly (not update_context_and_get).
+        // This test verifies that path works correctly (Issue #807: full branch coverage).
+        let sched = make_test_scheduler(0); // 0 = unlimited, but user provides None
+        let job_id = sched
+            .dispatch_job("user1", "test", "desc", None) // None metadata
+            .await
+            .unwrap(); // safety: test code
+
+        let ctx = sched.context_manager.get_context(job_id).await.unwrap(); // safety: test code
+        // No metadata was set, should have default empty metadata
+        assert!(ctx.metadata.is_null() || ctx.metadata == serde_json::json!({})); // safety: test code
+        // No user tokens AND unlimited config means max_tokens stays at default
+        assert_eq!(ctx.max_tokens, 0, "unlimited config"); // safety: test code
+    }
+
     #[test]
     fn test_scheduler_creation() {
         // Would need to mock dependencies for proper testing

src/channels/web/handlers/routines.rs

@@ -112,12 +112,16 @@ pub async fn routines_detail_handler(
             job_id: run.job_id,
         })
         .collect();
+    let routine_info = RoutineInfo::from_routine(&routine);
 
     Ok(Json(RoutineDetailResponse {
         id: routine.id,
         name: routine.name.clone(),
         description: routine.description.clone(),
         enabled: routine.enabled,
+        trigger_type: routine_info.trigger_type,
+        trigger_raw: routine_info.trigger_raw,
+        trigger_summary: routine_info.trigger_summary,
         trigger: serde_json::to_value(&routine.trigger).unwrap_or_default(),
         action: serde_json::to_value(&routine.action).unwrap_or_default(),
         guardrails: serde_json::to_value(&routine.guardrails).unwrap_or_default(),

src/channels/web/openai_compat.rs

@@ -419,6 +419,44 @@ fn parse_stop(val: &serde_json::Value) -> Option<Vec<String>> {
     }
 }
 
+fn build_completion_request(
+    req: &OpenAiChatRequest,
+    messages: Vec<ChatMessage>,
+) -> CompletionRequest {
+    let mut comp_req = CompletionRequest::new(messages).with_model(req.model.clone());
+    if let Some(t) = req.temperature {
+        comp_req = comp_req.with_temperature(t);
+    }
+    if let Some(mt) = req.max_tokens {
+        comp_req = comp_req.with_max_tokens(mt);
+    }
+    if let Some(stops) = req.stop.as_ref().and_then(parse_stop) {
+        comp_req.stop_sequences = Some(stops);
+    }
+    comp_req
+}
+
+fn build_tool_request(
+    req: &OpenAiChatRequest,
+    messages: Vec<ChatMessage>,
+) -> ToolCompletionRequest {
+    let tools = convert_tools(req.tools.as_deref().unwrap_or(&[]));
+    let mut tool_req = ToolCompletionRequest::new(messages, tools).with_model(req.model.clone());
+    if let Some(t) = req.temperature {
+        tool_req = tool_req.with_temperature(t);
+    }
+    if let Some(mt) = req.max_tokens {
+        tool_req = tool_req.with_max_tokens(mt);
+    }
+    if let Some(stops) = req.stop.as_ref().and_then(parse_stop) {
+        tool_req = tool_req.with_stop_sequences(stops);
+    }
+    if let Some(choice) = req.tool_choice.as_ref().and_then(normalize_tool_choice) {
+        tool_req = tool_req.with_tool_choice(choice);
+    }
+    tool_req
+}
+
 // ---------------------------------------------------------------------------
 // Handlers
 // ---------------------------------------------------------------------------
@@ -476,19 +514,7 @@ pub async fn chat_completions_handler(
     let created = unix_timestamp();
 
     if has_tools {
-        let tools = convert_tools(req.tools.as_deref().unwrap_or(&[]));
-        let mut tool_req = ToolCompletionRequest::new(messages, tools).with_model(req.model);
-        if let Some(t) = req.temperature {
-            tool_req = tool_req.with_temperature(t);
-        }
-        if let Some(mt) = req.max_tokens {
-            tool_req = tool_req.with_max_tokens(mt);
-        }
-        if let Some(ref tc) = req.tool_choice
-            && let Some(choice) = normalize_tool_choice(tc)
-        {
-            tool_req = tool_req.with_tool_choice(choice);
-        }
+        let tool_req = build_tool_request(&req, messages);
 
         let resp = llm
             .complete_with_tools(tool_req)
@@ -527,16 +553,7 @@ pub async fn chat_completions_handler(
 
         Ok(Json(response).into_response())
     } else {
-        let mut comp_req = CompletionRequest::new(messages).with_model(req.model);
-        if let Some(t) = req.temperature {
-            comp_req = comp_req.with_temperature(t);
-        }
-        if let Some(mt) = req.max_tokens {
-            comp_req = comp_req.with_max_tokens(mt);
-        }
-        if let Some(ref stop_val) = req.stop {
-            comp_req.stop_sequences = parse_stop(stop_val);
-        }
+        let comp_req = build_completion_request(&req, messages);
 
         let resp = llm.complete(comp_req).await.map_err(map_llm_error)?;
         let model_name = llm.effective_model_name(Some(requested_model.as_str()));
@@ -596,35 +613,14 @@ async fn handle_streaming(
     }
 
     let llm_result = if has_tools {
-        let tools = convert_tools(req.tools.as_deref().unwrap_or(&[]));
-        let mut tool_req = ToolCompletionRequest::new(messages, tools).with_model(req.model);
-        if let Some(t) = req.temperature {
-            tool_req = tool_req.with_temperature(t);
-        }
-        if let Some(mt) = req.max_tokens {
-            tool_req = tool_req.with_max_tokens(mt);
-        }
-        if let Some(ref tc) = req.tool_choice
-            && let Some(choice) = normalize_tool_choice(tc)
-        {
-            tool_req = tool_req.with_tool_choice(choice);
-        }
+        let tool_req = build_tool_request(&req, messages);
         LlmResult::WithTools(
             llm.complete_with_tools(tool_req)
                 .await
                 .map_err(map_llm_error)?,
         )
     } else {
-        let mut comp_req = CompletionRequest::new(messages).with_model(req.model);
-        if let Some(t) = req.temperature {
-            comp_req = comp_req.with_temperature(t);
-        }
-        if let Some(mt) = req.max_tokens {
-            comp_req = comp_req.with_max_tokens(mt);
-        }
-        if let Some(ref stop_val) = req.stop {
-            comp_req.stop_sequences = parse_stop(stop_val);
-        }
+        let comp_req = build_completion_request(&req, messages);
         LlmResult::Simple(llm.complete(comp_req).await.map_err(map_llm_error)?)
     };
     let model_name = llm.effective_model_name(Some(requested_model.as_str()));

src/channels/web/server.rs

@@ -2346,12 +2346,16 @@ async fn routines_detail_handler(
             job_id: run.job_id,
         })
         .collect();
+    let routine_info = RoutineInfo::from_routine(&routine);
 
     Ok(Json(RoutineDetailResponse {
         id: routine.id,
         name: routine.name.clone(),
         description: routine.description.clone(),
         enabled: routine.enabled,
+        trigger_type: routine_info.trigger_type,
+        trigger_raw: routine_info.trigger_raw,
+        trigger_summary: routine_info.trigger_summary,
         trigger: serde_json::to_value(&routine.trigger).unwrap_or_default(),
         action: serde_json::to_value(&routine.action).unwrap_or_default(),
         guardrails: serde_json::to_value(&routine.guardrails).unwrap_or_default(),

src/channels/web/static/app.js

@@ -3535,10 +3535,13 @@ function renderRoutinesList(routines) {
 
     const toggleLabel = r.enabled ? 'Disable' : 'Enable';
     const toggleClass = r.enabled ? 'btn-cancel' : 'btn-restart';
+    const triggerTitle = (r.trigger_type === 'cron' && r.trigger_raw)
+      ? ' title="' + escapeHtml(r.trigger_raw) + '"'
+      : '';
 
     return '<tr class="routine-row" data-action="open-routine" data-id="' + escapeHtml(r.id) + '">'
       + '<td>' + escapeHtml(r.name) + '</td>'
-      + '<td>' + escapeHtml(r.trigger_summary) + '</td>'
+      + '<td' + triggerTitle + '>' + escapeHtml(r.trigger_summary) + '</td>'
       + '<td>' + escapeHtml(r.action_type) + '</td>'
       + '<td>' + formatRelativeTime(r.last_run_at) + '</td>'
       + '<td>' + formatRelativeTime(r.next_fire_at) + '</td>'
@@ -3606,8 +3609,23 @@ function renderRoutineDetail(routine) {
   }
 
   // Trigger config
-  html += '<div class="job-description"><h3>Trigger</h3>'
-    + '<pre class="action-json">' + escapeHtml(JSON.stringify(routine.trigger, null, 2)) + '</pre></div>';
+  if (routine.trigger_type === 'cron') {
+    const summary = routine.trigger_summary || 'cron';
+    const raw = routine.trigger_raw || '';
+    const timezone = routine.trigger && routine.trigger.timezone ? String(routine.trigger.timezone) : '';
+    html += '<div class="job-description"><h3>Trigger</h3>'
+      + '<div class="job-description-body"><strong>' + escapeHtml(summary) + '</strong></div>';
+    if (raw) {
+      html += '<div class="job-meta-item">'
+        + '<span class="job-meta-label">Raw</span>'
+        + '<span class="job-meta-value">' + escapeHtml(raw + (timezone ? ' (' + timezone + ')' : '')) + '</span>'
+        + '</div>';
+    }
+    html += '</div>';
+  } else {
+    html += '<div class="job-description"><h3>Trigger</h3>'
+      + '<pre class="action-json">' + escapeHtml(JSON.stringify(routine.trigger, null, 2)) + '</pre></div>';
+  }
 
   // Action config
   html += '<div class="job-description"><h3>Action</h3>'

src/channels/web/types.rs

@@ -735,6 +735,7 @@ pub struct RoutineInfo {
     pub description: String,
     pub enabled: bool,
     pub trigger_type: String,
+    pub trigger_raw: String,
     pub trigger_summary: String,
     pub action_type: String,
     pub last_run_at: Option<String>,
@@ -747,25 +748,34 @@ pub struct RoutineInfo {
 impl RoutineInfo {
     /// Convert a `Routine` to the trimmed `RoutineInfo` for list display.
     pub fn from_routine(r: &crate::agent::routine::Routine) -> Self {
-        let (trigger_type, trigger_summary) = match &r.trigger {
-            crate::agent::routine::Trigger::Cron { schedule, .. } => {
-                ("cron".to_string(), format!("cron: {}", schedule))
-            }
+        let (trigger_type, trigger_raw, trigger_summary) = match &r.trigger {
+            crate::agent::routine::Trigger::Cron { schedule, timezone } => (
+                "cron".to_string(),
+                schedule.clone(),
+                crate::agent::routine::describe_cron(schedule, timezone.as_deref()),
+            ),
             crate::agent::routine::Trigger::Event {
                 pattern, channel, ..
             } => {
                 let ch = channel.as_deref().unwrap_or("any");
-                ("event".to_string(), format!("on {} /{}/", ch, pattern))
+                (
+                    "event".to_string(),
+                    String::new(),
+                    format!("on {} /{}/", ch, pattern),
+                )
             }
             crate::agent::routine::Trigger::SystemEvent {
                 source, event_type, ..
             } => (
                 "system_event".to_string(),
+                String::new(),
                 format!("event: {}.{}", source, event_type),
             ),
-            crate::agent::routine::Trigger::Manual => {
-                ("manual".to_string(), "manual only".to_string())
-            }
+            crate::agent::routine::Trigger::Manual => (
+                "manual".to_string(),
+                String::new(),
+                "manual only".to_string(),
+            ),
         };
 
         let action_type = match &r.action {
@@ -787,6 +797,7 @@ impl RoutineInfo {
             description: r.description.clone(),
             enabled: r.enabled,
             trigger_type,
+            trigger_raw,
             trigger_summary,
             action_type: action_type.to_string(),
             last_run_at: r.last_run_at.map(|dt| dt.to_rfc3339()),
@@ -818,6 +829,9 @@ pub struct RoutineDetailResponse {
     pub name: String,
     pub description: String,
     pub enabled: bool,
+    pub trigger_type: String,
+    pub trigger_raw: String,
+    pub trigger_summary: String,
     pub trigger: serde_json::Value,
     pub action: serde_json::Value,
     pub guardrails: serde_json::Value,

src/channels/webhook_server.rs

@@ -139,12 +139,19 @@ impl WebhookServer {
         self.config.addr
     }
 
+    /// Take ownership of shutdown primitives so callers can perform async
+    /// shutdown work without holding external locks around this server.
+    pub fn begin_shutdown(&mut self) -> (Option<oneshot::Sender<()>>, Option<JoinHandle<()>>) {
+        (self.shutdown_tx.take(), self.handle.take())
+    }
+
     /// Signal graceful shutdown and wait for the server task to finish.
     pub async fn shutdown(&mut self) {
-        if let Some(tx) = self.shutdown_tx.take() {
+        let (shutdown_tx, handle) = self.begin_shutdown();
+        if let Some(tx) = shutdown_tx {
             let _ = tx.send(());
         }
-        if let Some(handle) = self.handle.take() {
+        if let Some(handle) = handle {
             let _ = handle.await;
         }
     }
@@ -269,6 +276,35 @@ mod tests {
         server.shutdown().await;
     }
 
+    #[tokio::test]
+    async fn test_begin_shutdown_takes_handles_for_lock_free_shutdown() {
+        let addr = SocketAddr::from((std::net::Ipv4Addr::LOCALHOST, 0));
+        let mut server = WebhookServer::new(WebhookServerConfig { addr });
+
+        let test_router = axum::Router::new().route(
+            "/health",
+            axum::routing::get(|| async { Json(json!({"status": "ok"})) }),
+        );
+        server.add_routes(test_router);
+        server.start().await.expect("Failed to start server"); // safety: test assertion for setup precondition
+
+        let (shutdown_tx, handle) = server.begin_shutdown();
+        assert!(shutdown_tx.is_some(), "shutdown sender should be available"); // safety: test assertion for expected server state
+        assert!(handle.is_some(), "server handle should be available"); // safety: test assertion for expected server state
+
+        // begin_shutdown() should leave no handles behind on the server.
+        let (shutdown_tx2, handle2) = server.begin_shutdown();
+        assert!(shutdown_tx2.is_none(), "shutdown sender should be consumed"); // safety: test assertion for postcondition
+        assert!(handle2.is_none(), "server handle should be consumed"); // safety: test assertion for postcondition
+
+        if let Some(tx) = shutdown_tx {
+            let _ = tx.send(());
+        }
+        if let Some(handle) = handle {
+            let _ = handle.await;
+        }
+    }
+
     #[tokio::test]
     async fn test_restart_with_addr_rollback_on_bind_failure() {
         use std::net::TcpListener as StdTcpListener;

src/context/manager.rs

@@ -87,6 +87,28 @@ impl ContextManager {
         Ok(f(context))
     }
 
+    /// Atomically update a job context and return the updated context.
+    ///
+    /// This method holds the write lock for the entire update-and-read sequence,
+    /// preventing concurrent workers from interleaving modifications between the
+    /// update and the subsequent read (Issue #807: non-transactional context updates).
+    /// Use this when you need to update context and immediately persist it to DB.
+    pub async fn update_context_and_get<F>(
+        &self,
+        job_id: Uuid,
+        f: F,
+    ) -> Result<JobContext, JobError>
+    where
+        F: FnOnce(&mut JobContext),
+    {
+        let mut contexts = self.contexts.write().await;
+        let context = contexts
+            .get_mut(&job_id)
+            .ok_or(JobError::NotFound { id: job_id })?;
+        f(context);
+        Ok(context.clone())
+    }
+
     /// Get job memory.
     pub async fn get_memory(&self, job_id: Uuid) -> Result<Memory, JobError> {
         self.memories
@@ -877,4 +899,70 @@ mod tests {
 
         assert_eq!(manager.all_jobs().await.len(), 10);
     }
+
+    #[tokio::test]
+    async fn update_context_and_get_atomicity_regression_issue_807() {
+        // Regression test for Issue #807: non-transactional context updates.
+        // Verify that update_context_and_get returns the exact state that was set,
+        // without allowing concurrent workers to interleave modifications.
+        let manager = std::sync::Arc::new(ContextManager::new(100));
+        let job_id = manager
+            .create_job("Atomicity Test", "verify no race condition")
+            .await
+            .unwrap(); // safety: test code
+
+        // Update and get atomically, setting metadata
+        let metadata = serde_json::json!({ "priority": "high", "user_id": 42 });
+        let returned_ctx = manager
+            .update_context_and_get(job_id, |ctx| {
+                ctx.metadata = metadata.clone();
+                ctx.max_tokens = 5000;
+            })
+            .await
+            .unwrap(); // safety: test code
+
+        // Verify the returned context has the exact updates we set
+        assert_eq!(returned_ctx.metadata, metadata); // safety: test code
+        assert_eq!(returned_ctx.max_tokens, 5000); // safety: test code
+
+        // Verify a fresh get returns the same state
+        let fresh_ctx = manager.get_context(job_id).await.unwrap(); // safety: test code
+        assert_eq!(fresh_ctx.metadata, metadata); // safety: test code
+        assert_eq!(fresh_ctx.max_tokens, 5000); // safety: test code
+    }
+
+    #[tokio::test]
+    async fn update_context_and_get_no_concurrent_interleave() {
+        // Verify that concurrent updates cannot interleave during update_context_and_get.
+        // If the lock were released too early, a concurrent state transition could
+        // get mixed into the returned context.
+        let manager = std::sync::Arc::new(ContextManager::new(100));
+        let job_id = manager
+            .create_job("Concurrent Race Test", "ensure atomicity")
+            .await
+            .unwrap(); // safety: test code
+
+        let metadata = serde_json::json!({ "test": "race_condition" });
+        let metadata_clone = metadata.clone();
+
+        // Spawn a task that will update_context_and_get
+        let mgr1 = std::sync::Arc::clone(&manager);
+        let returned_ctx_handle = tokio::spawn(async move {
+            mgr1.update_context_and_get(job_id, |ctx| {
+                ctx.metadata = metadata_clone;
+                ctx.max_tokens = 3000;
+            })
+            .await
+        });
+
+        // The returned context should have *only* the metadata update, not any
+        // concurrent state transitions that might happen during the operation.
+        let returned_ctx = returned_ctx_handle.await.unwrap().unwrap(); // safety: test code
+
+        // Verify atomicity: returned context has the metadata we set
+        assert_eq!(returned_ctx.metadata, metadata); // safety: test code
+        assert_eq!(returned_ctx.max_tokens, 3000); // safety: test code
+        // And it's in the initial state (Pending), not modified by concurrent workers
+        assert_eq!(returned_ctx.state, crate::context::JobState::Pending); // safety: test code
+    }
 }

src/db/libsql/routines.rs

@@ -1,13 +1,15 @@
 //! Routine-related RoutineStore implementation for LibSqlBackend.
 
+use std::collections::{HashMap, HashSet};
+
 use async_trait::async_trait;
 use chrono::{DateTime, Utc};
 use libsql::params;
 use uuid::Uuid;
 
 use super::{
-    LibSqlBackend, ROUTINE_COLUMNS, ROUTINE_RUN_COLUMNS, fmt_opt_ts, fmt_ts, get_i64, opt_text,
-    opt_text_owned, row_to_routine_libsql, row_to_routine_run_libsql,
+    LibSqlBackend, ROUTINE_COLUMNS, ROUTINE_RUN_COLUMNS, fmt_opt_ts, fmt_ts, get_i64, get_text,
+    opt_text, opt_text_owned, row_to_routine_libsql, row_to_routine_run_libsql,
 };
 use crate::agent::routine::{Routine, RoutineRun, RunStatus};
 use crate::db::RoutineStore;
@@ -409,6 +411,57 @@ impl RoutineStore for LibSqlBackend {
         }
     }
 
+    async fn count_running_routine_runs_batch(
+        &self,
+        routine_ids: &[Uuid],
+    ) -> Result<HashMap<Uuid, i64>, DatabaseError> {
+        if routine_ids.is_empty() {
+            return Ok(HashMap::new());
+        }
+
+        let mut counts = HashMap::new();
+        let conn = self.connect().await?;
+
+        // Query all running routines and filter in memory
+        // This is simpler for libSQL than building dynamic parameter lists
+        let mut rows = conn
+            .query(
+                "SELECT routine_id, COUNT(*) as cnt FROM routine_runs
+                 WHERE status = 'running'
+                 GROUP BY routine_id",
+                params![],
+            )
+            .await
+            .map_err(|e| {
+                DatabaseError::Query(format!("Failed to batch count running routines: {}", e))
+            })?;
+
+        let routine_id_set: HashSet<Uuid> = routine_ids.iter().copied().collect();
+
+        while let Some(row) = rows
+            .next()
+            .await
+            .map_err(|e| DatabaseError::Query(e.to_string()))?
+        {
+            let id_str: String = get_text(&row, 0);
+            let id = Uuid::parse_str(&id_str)
+                .map_err(|e| DatabaseError::Query(format!("Invalid routine UUID: {}", e)))?;
+
+            // Only include if this routine ID was requested
+            if routine_id_set.contains(&id) {
+                let cnt: i64 = get_i64(&row, 1);
+                counts.insert(id, cnt);
+            }
+        }
+
+        // Ensure all requested IDs are in the map (defaults to 0 for no running runs)
+        for id in routine_ids {
+            counts.entry(*id).or_insert(0);
+        }
+
+        Ok(counts)
+    }
+
     async fn link_routine_run_to_job(
         &self,
         run_id: Uuid,

src/db/mod.rs

@@ -387,6 +387,10 @@ pub trait RoutineStore: Send + Sync {
         limit: i64,
     ) -> Result<Vec<RoutineRun>, DatabaseError>;
     async fn count_running_routine_runs(&self, routine_id: Uuid) -> Result<i64, DatabaseError>;
+    async fn count_running_routine_runs_batch(
+        &self,
+        routine_ids: &[Uuid],
+    ) -> Result<HashMap<Uuid, i64>, DatabaseError>;
     async fn link_routine_run_to_job(
         &self,
         run_id: Uuid,

src/db/postgres.rs

@@ -487,6 +487,15 @@ impl RoutineStore for PgBackend {
         self.store.count_running_routine_runs(routine_id).await
     }
 
+    async fn count_running_routine_runs_batch(
+        &self,
+        routine_ids: &[Uuid],
+    ) -> Result<std::collections::HashMap<Uuid, i64>, DatabaseError> {
+        self.store
+            .count_running_routine_runs_batch(routine_ids)
+            .await
+    }
+
     async fn link_routine_run_to_job(
         &self,
         run_id: Uuid,

src/extensions/manager.rs

@@ -304,6 +304,34 @@ impl ExtensionManager {
         *self.relay_channel_manager.write().await = Some(channel_manager);
     }
 
+    async fn current_channel_owner_id(&self, name: &str) -> Option<i64> {
+        {
+            let rt_guard = self.channel_runtime.read().await;
+            if let Some(owner_id) = rt_guard
+                .as_ref()
+                .and_then(|rt| rt.wasm_channel_owner_ids.get(name).copied())
+            {
+                return Some(owner_id);
+            }
+        }
+
+        let store = self.store.as_ref()?;
+        let key = format!("channels.wasm_channel_owner_ids.{name}");
+        match store.get_setting(&self.user_id, &key).await {
+            Ok(Some(serde_json::Value::Number(n))) => n.as_i64(),
+            Ok(Some(serde_json::Value::String(s))) => s.parse::<i64>().ok(),
+            Ok(Some(_)) | Ok(None) => None,
+            Err(e) => {
+                tracing::debug!(
+                    channel = %name,
+                    error = %e,
+                    "Failed to read persisted wasm channel owner id"
+                );
+                None
+            }
+        }
+    }
+
     /// Check if a channel name corresponds to a relay extension (has stored stream token).
     pub async fn is_relay_channel(&self, name: &str) -> bool {
         self.secrets
@@ -2980,13 +3008,7 @@ impl ExtensionManager {
 
         // Verify runtime infrastructure is available and clone Arcs so we don't
         // hold the RwLock guard across awaits.
-        let (
-            channel_runtime,
-            channel_manager,
-            pairing_store,
-            wasm_channel_router,
-            wasm_channel_owner_ids,
-        ) = {
+        let (channel_runtime, channel_manager, pairing_store, wasm_channel_router) = {
             let rt_guard = self.channel_runtime.read().await;
             let rt = rt_guard.as_ref().ok_or_else(|| {
                 ExtensionError::ActivationFailed("WASM channel runtime not configured".to_string())
@@ -2996,7 +3018,6 @@ impl ExtensionManager {
                 Arc::clone(&rt.channel_manager),
                 Arc::clone(&rt.pairing_store),
                 Arc::clone(&rt.wasm_channel_router),
-                rt.wasm_channel_owner_ids.clone(),
             )
         };
 
@@ -3067,7 +3088,7 @@ impl ExtensionManager {
                 );
             }
 
-            if let Some(&owner_id) = wasm_channel_owner_ids.get(channel_name.as_str()) {
+            if let Some(owner_id) = self.current_channel_owner_id(&channel_name).await {
                 config_updates.insert("owner_id".to_string(), serde_json::json!(owner_id));
             }
 
@@ -4744,6 +4765,126 @@ mod tests {
         )
     }
 
+    #[tokio::test]
+    async fn test_current_channel_owner_id_uses_runtime_state() -> Result<(), String> {
+        let manager = make_manager_with_temp_dirs();
+        if manager.current_channel_owner_id("telegram").await.is_some() {
+            return Err("expected no owner id for telegram before runtime setup".to_string());
+        }
+
+        let channels = Arc::new(crate::channels::ChannelManager::new());
+        let runtime = Arc::new(
+            crate::channels::wasm::WasmChannelRuntime::new(
+                crate::channels::wasm::WasmChannelRuntimeConfig::default(),
+            )
+            .map_err(|e| format!("runtime init failed: {e}"))?,
+        );
+        let pairing_store = Arc::new(crate::pairing::PairingStore::new());
+        let router = Arc::new(crate::channels::wasm::WasmChannelRouter::new());
+        let mut owner_ids = std::collections::HashMap::new();
+        owner_ids.insert("telegram".to_string(), 12345_i64);
+
+        manager
+            .set_channel_runtime(channels, runtime, pairing_store, router, owner_ids)
+            .await;
+
+        if manager.current_channel_owner_id("telegram").await != Some(12345_i64) {
+            return Err("expected runtime owner id fast-path for telegram".to_string());
+        }
+        if manager.current_channel_owner_id("slack").await.is_some() {
+            return Err("expected no owner id for slack".to_string());
+        }
+
+        Ok(())
+    }
+
+    #[cfg(feature = "libsql")]
+    #[tokio::test]
+    async fn test_current_channel_owner_id_uses_store_fallback() -> Result<(), String> {
+        use crate::db::{Database, SettingsStore};
+
+        let dir = tempfile::tempdir().map_err(|e| format!("tempdir failed: {e}"))?;
+        let db_path = dir.path().join("owner-id.db");
+
+        let db = Arc::new(
+            crate::db::libsql::LibSqlBackend::new_local(&db_path)
+                .await
+                .map_err(|e| format!("create local libsql backend failed: {e}"))?,
+        );
+        db.run_migrations()
+            .await
+            .map_err(|e| format!("run libsql migrations failed: {e}"))?;
+
+        let tools_dir = dir.path().join("tools");
+        let channels_dir = dir.path().join("channels");
+        std::fs::create_dir_all(&tools_dir).ok();
+        std::fs::create_dir_all(&channels_dir).ok();
+
+        use crate::secrets::{InMemorySecretsStore, SecretsCrypto};
+        use crate::testing::credentials::TEST_CRYPTO_KEY;
+        use crate::tools::ToolRegistry;
+        use crate::tools::mcp::process::McpProcessManager;
+        use crate::tools::mcp::session::McpSessionManager;
+
+        let master_key = secrecy::SecretString::from(TEST_CRYPTO_KEY.to_string());
+        let crypto = Arc::new(
+            SecretsCrypto::new(master_key)
+                .map_err(|e| format!("create secrets crypto failed: {e}"))?,
+        );
+
+        let manager = ExtensionManager::new(
+            Arc::new(McpSessionManager::new()),
+            Arc::new(McpProcessManager::new()),
+            Arc::new(InMemorySecretsStore::new(crypto)),
+            Arc::new(ToolRegistry::new()),
+            None,
+            None,
+            tools_dir,
+            channels_dir,
+            None,
+            "test".to_string(),
+            Some(db.clone() as Arc<dyn crate::db::Database>),
+            Vec::new(),
+        );
+
+        if manager.current_channel_owner_id("telegram").await.is_some() {
+            return Err("expected no owner id before settings seed".to_string());
+        }
+
+        db.set_setting(
+            "test",
+            "channels.wasm_channel_owner_ids.telegram",
+            &serde_json::json!(54321_i64),
+        )
+        .await
+        .map_err(|e| format!("persist owner id in settings failed: {e}"))?;
+
+        if manager.current_channel_owner_id("telegram").await != Some(54321_i64) {
+            return Err("expected store fallback owner id for telegram".to_string());
+        }
+
+        let channels = Arc::new(crate::channels::ChannelManager::new());
+        let runtime = Arc::new(
+            crate::channels::wasm::WasmChannelRuntime::new(
+                crate::channels::wasm::WasmChannelRuntimeConfig::default(),
+            )
+            .map_err(|e| format!("runtime init failed: {e}"))?,
+        );
+        let pairing_store = Arc::new(crate::pairing::PairingStore::new());
+        let router = Arc::new(crate::channels::wasm::WasmChannelRouter::new());
+        let mut owner_ids = std::collections::HashMap::new();
+        owner_ids.insert("telegram".to_string(), 12345_i64);
+        manager
+            .set_channel_runtime(channels, runtime, pairing_store, router, owner_ids)
+            .await;
+
+        if manager.current_channel_owner_id("telegram").await != Some(12345_i64) {
+            return Err("expected runtime fast-path owner id precedence".to_string());
+        }
+
+        Ok(())
+    }
+
     // ── resolve_env_credentials tests ────────────────────────────────────
 
     #[test]

src/history/store.rs

@@ -1,5 +1,8 @@
 //! PostgreSQL store for persisting agent data.
 
+#[cfg(feature = "postgres")]
+use std::collections::HashMap;
+
 use chrono::{DateTime, Utc};
 #[cfg(feature = "postgres")]
 use deadpool_postgres::{Config, Pool};
@@ -1294,6 +1297,42 @@ impl Store {
         Ok(row.get("cnt"))
     }
 
+    /// Batch-load concurrent run counts for multiple routines in a single query.
+    /// Returns a map where missing routine IDs default to 0.
+    #[cfg(feature = "postgres")]
+    pub async fn count_running_routine_runs_batch(
+        &self,
+        routine_ids: &[Uuid],
+    ) -> Result<HashMap<Uuid, i64>, DatabaseError> {
+        if routine_ids.is_empty() {
+            return Ok(HashMap::new());
+        }
+
+        let conn = self.conn().await?;
+        let rows = conn
+            .query(
+                "SELECT routine_id, COUNT(*) as cnt FROM routine_runs
+                 WHERE routine_id = ANY($1) AND status = 'running'
+                 GROUP BY routine_id",
+                &[&routine_ids],
+            )
+            .await?;
+
+        let mut counts = HashMap::new();
+        for row in rows {
+            let id: Uuid = row.get("routine_id");
+            let cnt: i64 = row.get("cnt");
+            counts.insert(id, cnt);
+        }
+
+        // Ensure all requested IDs are in the map (defaults to 0 for no running runs)
+        for id in routine_ids {
+            counts.entry(*id).or_insert(0);
+        }
+
+        Ok(counts)
+    }
+
     /// Link a routine run to a dispatched job.
     pub async fn link_routine_run_to_job(
         &self,

src/llm/bedrock.rs

@@ -176,8 +176,11 @@ impl LlmProvider for BedrockProvider {
             builder = builder.tool_config(tc);
         }
 
-        if let Some(config) = build_inference_config(request.temperature, request.max_tokens, None)
-        {
+        if let Some(config) = build_inference_config(
+            request.temperature,
+            request.max_tokens,
+            request.stop_sequences.as_deref(),
+        ) {
             builder = builder.inference_config(config);
         }
 

src/llm/nearai_chat.rs

@@ -475,6 +475,7 @@ impl LlmProvider for NearAiChatProvider {
             messages,
             temperature: req.temperature,
             max_tokens: req.max_tokens,
+            stop: req.stop_sequences,
             tools: None,
             tool_choice: None,
         };
@@ -554,6 +555,7 @@ impl LlmProvider for NearAiChatProvider {
             messages,
             temperature: req.temperature,
             max_tokens: req.max_tokens,
+            stop: req.stop_sequences,
             tools: if tools.is_empty() { None } else { Some(tools) },
             tool_choice: req.tool_choice,
         };
@@ -680,6 +682,8 @@ struct ChatCompletionRequest {
     #[serde(skip_serializing_if = "Option::is_none")]
     max_tokens: Option<u32>,
     #[serde(skip_serializing_if = "Option::is_none")]
+    stop: Option<Vec<String>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
     tools: Option<Vec<ChatCompletionTool>>,
     #[serde(skip_serializing_if = "Option::is_none")]
     tool_choice: Option<String>,
@@ -1666,6 +1670,7 @@ mod tests {
             }],
             temperature: None,
             max_tokens: None,
+            stop: None,
             tools: None,
             tool_choice: None,
         };
@@ -1687,6 +1692,7 @@ mod tests {
             messages: vec![],
             temperature: Some(0.7),
             max_tokens: Some(1024),
+            stop: None,
             tools: Some(vec![ChatCompletionTool {
                 tool_type: "function".to_string(),
                 function: ChatCompletionFunction {

src/llm/provider.rs

@@ -251,6 +251,7 @@ pub struct ToolCompletionRequest {
     pub model: Option<String>,
     pub max_tokens: Option<u32>,
     pub temperature: Option<f32>,
+    pub stop_sequences: Option<Vec<String>>,
     /// How to handle tool use: "auto", "required", or "none".
     pub tool_choice: Option<String>,
     /// Opaque metadata passed through to the provider (e.g. thread_id for chaining).
@@ -266,6 +267,7 @@ impl ToolCompletionRequest {
             model: None,
             max_tokens: None,
             temperature: None,
+            stop_sequences: None,
             tool_choice: None,
             metadata: std::collections::HashMap::new(),
         }
@@ -289,6 +291,12 @@ impl ToolCompletionRequest {
         self
     }
 
+    /// Set stop sequences.
+    pub fn with_stop_sequences(mut self, stop_sequences: Vec<String>) -> Self {
+        self.stop_sequences = Some(stop_sequences);
+        self
+    }
+
     /// Set tool choice mode.
     pub fn with_tool_choice(mut self, choice: impl Into<String>) -> Self {
         self.tool_choice = Some(choice.into());
@@ -504,8 +512,6 @@ pub fn strip_unsupported_completion_params(
 /// This is the single helper function used by all providers to remove
 /// parameters they don't support from tool calls, replacing duplicate stringly-typed logic.
 ///
-/// Note: Only `Temperature` and `MaxTokens` are supported in `ToolCompletionRequest`.
-/// `StopSequences` is only available in `CompletionRequest` and is not applicable to tool calls.
 pub fn strip_unsupported_tool_params(
     unsupported: &std::collections::HashSet<String>,
     req: &mut ToolCompletionRequest,
@@ -519,7 +525,9 @@ pub fn strip_unsupported_tool_params(
     if unsupported.contains(UnsupportedParam::MaxTokens.name()) {
         req.max_tokens = None;
     }
-    // Note: StopSequences is not a field in ToolCompletionRequest, so no action needed
+    if unsupported.contains(UnsupportedParam::StopSequences.name()) {
+        req.stop_sequences = None;
+    }
 }
 
 #[cfg(test)]
@@ -651,4 +659,17 @@ mod tests {
         assert!(messages[2].tool_call_id.is_none());
         assert!(messages[2].name.is_none());
     }
+
+    #[test]
+    fn test_strip_unsupported_tool_params_strips_stop_sequences() {
+        let mut unsupported = std::collections::HashSet::new();
+        unsupported.insert(UnsupportedParam::StopSequences.name().to_string());
+
+        let mut req = ToolCompletionRequest::new(vec![ChatMessage::user("hello")], vec![]);
+        req.stop_sequences = Some(vec!["STOP".to_string()]);
+
+        strip_unsupported_tool_params(&unsupported, &mut req);
+
+        assert!(req.stop_sequences.is_none()); // safety: test assertion for explicit strip behavior
+    }
 }

src/llm/response_cache.rs

@@ -548,6 +548,7 @@ mod tests {
             model: None,
             max_tokens: None,
             temperature: None,
+            stop_sequences: None,
             tool_choice: None,
             metadata: Default::default(),
         };

src/main.rs

@@ -920,7 +920,16 @@ async fn async_main() -> anyhow::Result<()> {
     }
 
     if let Some(ref ws_arc) = webhook_server {
-        ws_arc.lock().await.shutdown().await;
+        let (shutdown_tx, handle) = {
+            let mut ws = ws_arc.lock().await;
+            ws.begin_shutdown()
+        };
+        if let Some(tx) = shutdown_tx {
+            let _ = tx.send(());
+        }
+        if let Some(handle) = handle {
+            let _ = handle.await;
+        }
     }
 
     if let Some(tunnel) = active_tunnel {

src/orchestrator/api.rs

@@ -176,6 +176,7 @@ async fn llm_complete_with_tools(
         model: req.model,
         max_tokens: req.max_tokens,
         temperature: req.temperature,
+        stop_sequences: req.stop_sequences,
         tool_choice: req.tool_choice,
         metadata: std::collections::HashMap::new(),
     };

src/worker/api.rs

@@ -65,6 +65,7 @@ pub struct ProxyToolCompletionRequest {
     pub model: Option<String>,
     pub max_tokens: Option<u32>,
     pub temperature: Option<f32>,
+    pub stop_sequences: Option<Vec<String>>,
     pub tool_choice: Option<String>,
 }
 
@@ -251,6 +252,7 @@ impl WorkerHttpClient {
             model: request.model.clone(),
             max_tokens: request.max_tokens,
             temperature: request.temperature,
+            stop_sequences: request.stop_sequences.clone(),
             tool_choice: request.tool_choice.clone(),
         };
 

tests/batch_query_tests.rs

@@ -0,0 +1,509 @@
+//! Tests for batch loading routine concurrent counts (N+1 query fix).
+//!
+//! Verifies:
+//! 1. Batch query returns correct counts for multiple routines
+//! 2. Concurrent limit enforcement uses batch counts correctly
+
+#[cfg(feature = "libsql")]
+mod tests {
+    use std::sync::Arc;
+
+    use chrono::Utc;
+    use uuid::Uuid;
+
+    use ironclaw::agent::routine::{
+        Routine, RoutineAction, RoutineGuardrails, RoutineRun, RunStatus, Trigger,
+    };
+    use ironclaw::db::Database;
+
+    async fn create_test_db() -> (Arc<dyn Database>, tempfile::TempDir) {
+        use ironclaw::db::libsql::LibSqlBackend;
+
+        let temp_dir = tempfile::tempdir().expect("tempdir");
+        let db_path = temp_dir.path().join("test.db");
+        let backend = LibSqlBackend::new_local(&db_path)
+            .await
+            .expect("LibSqlBackend");
+        backend.run_migrations().await.expect("migrations");
+        let db: Arc<dyn Database> = Arc::new(backend);
+        (db, temp_dir)
+    }
+
+    // -----------------------------------------------------------------------
+    // Test 1: Batch query returns correct counts for multiple routines
+    // -----------------------------------------------------------------------
+
+    #[tokio::test]
+    async fn batch_query_empty_list() {
+        let (db, _tmp) = create_test_db().await;
+        let counts = db
+            .count_running_routine_runs_batch(&[])
+            .await
+            .expect("batch query should not fail");
+        assert!(counts.is_empty(), "Empty input should return empty map");
+    }
+
+    #[tokio::test]
+    async fn batch_query_single_routine() {
+        let (db, _tmp) = create_test_db().await;
+        let routine_id = Uuid::new_v4();
+
+        // Create routine
+        let routine = Routine {
+            id: routine_id,
+            name: "test-routine".to_string(),
+            description: "Test".to_string(),
+            user_id: "default".to_string(),
+            enabled: true,
+            trigger: Trigger::Cron {
+                schedule: "* * * * *".to_string(),
+                timezone: None,
+            },
+            action: RoutineAction::Lightweight {
+                prompt: "test".to_string(),
+                context_paths: vec![],
+                max_tokens: 1000,
+                use_tools: false,
+                max_tool_rounds: 3,
+            },
+            guardrails: RoutineGuardrails {
+                cooldown: std::time::Duration::from_secs(0),
+                max_concurrent: 5,
+                dedup_window: None,
+            },
+            notify: Default::default(),
+            last_run_at: None,
+            next_fire_at: None,
+            run_count: 0,
+            consecutive_failures: 0,
+            state: serde_json::json!({}),
+            created_at: Utc::now(),
+            updated_at: Utc::now(),
+        };
+        db.create_routine(&routine).await.expect("create routine");
+
+        // Create 3 running runs
+        for _ in 0..3 {
+            let run = RoutineRun {
+                id: Uuid::new_v4(),
+                routine_id,
+                trigger_type: "cron".to_string(),
+                trigger_detail: None,
+                started_at: Utc::now(),
+                completed_at: None,
+                status: RunStatus::Running,
+                result_summary: None,
+                tokens_used: None,
+                job_id: None,
+                created_at: Utc::now(),
+            };
+            db.create_routine_run(&run).await.expect("create run");
+        }
+
+        // Batch query for single routine
+        let counts = db
+            .count_running_routine_runs_batch(&[routine_id])
+            .await
+            .expect("batch query should work");
+
+        assert_eq!(counts.len(), 1, "Should return 1 routine");
+        assert_eq!(counts[&routine_id], 3, "Should count 3 running runs");
+    }
+
+    #[tokio::test]
+    async fn batch_query_multiple_routines_different_counts() {
+        let (db, _tmp) = create_test_db().await;
+
+        let r1 = Uuid::new_v4();
+        let r2 = Uuid::new_v4();
+        let r3 = Uuid::new_v4();
+
+        // Create 3 routines
+        for routine_id in [r1, r2, r3] {
+            let routine = Routine {
+                id: routine_id,
+                name: format!("routine-{}", routine_id),
+                description: "Test".to_string(),
+                user_id: "default".to_string(),
+                enabled: true,
+                trigger: Trigger::Cron {
+                    schedule: "* * * * *".to_string(),
+                    timezone: None,
+                },
+                action: RoutineAction::Lightweight {
+                    prompt: "test".to_string(),
+                    context_paths: vec![],
+                    max_tokens: 1000,
+                    use_tools: false,
+                    max_tool_rounds: 3,
+                },
+                guardrails: RoutineGuardrails {
+                    cooldown: std::time::Duration::from_secs(0),
+                    max_concurrent: 5,
+                    dedup_window: None,
+                },
+                notify: Default::default(),
+                last_run_at: None,
+                next_fire_at: None,
+                run_count: 0,
+                consecutive_failures: 0,
+                state: serde_json::json!({}),
+                created_at: Utc::now(),
+                updated_at: Utc::now(),
+            };
+            db.create_routine(&routine).await.expect("create routine");
+        }
+
+        // r1: 2 running
+        for _ in 0..2 {
+            let run = RoutineRun {
+                id: Uuid::new_v4(),
+                routine_id: r1,
+                trigger_type: "cron".to_string(),
+                trigger_detail: None,
+                started_at: Utc::now(),
+                completed_at: None,
+                status: RunStatus::Running,
+                result_summary: None,
+                tokens_used: None,
+                job_id: None,
+                created_at: Utc::now(),
+            };
+            db.create_routine_run(&run).await.expect("create run");
+        }
+
+        // r2: 1 running
+        let run = RoutineRun {
+            id: Uuid::new_v4(),
+            routine_id: r2,
+            trigger_type: "cron".to_string(),
+            trigger_detail: None,
+            started_at: Utc::now(),
+            completed_at: None,
+            status: RunStatus::Running,
+            result_summary: None,
+            tokens_used: None,
+            job_id: None,
+            created_at: Utc::now(),
+        };
+        db.create_routine_run(&run).await.expect("create run");
+
+        // r3: 0 running (but has 1 Ok result)
+        let run = RoutineRun {
+            id: Uuid::new_v4(),
+            routine_id: r3,
+            trigger_type: "cron".to_string(),
+            trigger_detail: None,
+            started_at: Utc::now(),
+            completed_at: Some(Utc::now()),
+            status: RunStatus::Ok,
+            result_summary: None,
+            tokens_used: None,
+            job_id: None,
+            created_at: Utc::now(),
+        };
+        db.create_routine_run(&run).await.expect("create run");
+
+        // Single batch query for all 3
+        let counts = db
+            .count_running_routine_runs_batch(&[r1, r2, r3])
+            .await
+            .expect("batch query should work");
+
+        assert_eq!(counts.len(), 3, "Should return 3 routines");
+        assert_eq!(counts[&r1], 2, "r1 should have 2 running");
+        assert_eq!(counts[&r2], 1, "r2 should have 1 running");
+        assert_eq!(
+            counts[&r3], 0,
+            "r3 should have 0 running (Ok status is not running)"
+        );
+    }
+
+    #[tokio::test]
+    async fn batch_query_missing_routines_default_to_zero() {
+        let (db, _tmp) = create_test_db().await;
+
+        let r1 = Uuid::new_v4();
+        let r2 = Uuid::new_v4();
+        let r3 = Uuid::new_v4(); // This one won't exist
+
+        // Only create r1
+        let routine = Routine {
+            id: r1,
+            name: "routine-1".to_string(),
+            description: "Test".to_string(),
+            user_id: "default".to_string(),
+            enabled: true,
+            trigger: Trigger::Cron {
+                schedule: "* * * * *".to_string(),
+                timezone: None,
+            },
+            action: RoutineAction::Lightweight {
+                prompt: "test".to_string(),
+                context_paths: vec![],
+                max_tokens: 1000,
+                use_tools: false,
+                max_tool_rounds: 3,
+            },
+            guardrails: RoutineGuardrails {
+                cooldown: std::time::Duration::from_secs(0),
+                max_concurrent: 5,
+                dedup_window: None,
+            },
+            notify: Default::default(),
+            last_run_at: None,
+            next_fire_at: None,
+            run_count: 0,
+            consecutive_failures: 0,
+            state: serde_json::json!({}),
+            created_at: Utc::now(),
+            updated_at: Utc::now(),
+        };
+        db.create_routine(&routine).await.expect("create routine");
+
+        // r1 has 1 running
+        let run = RoutineRun {
+            id: Uuid::new_v4(),
+            routine_id: r1,
+            trigger_type: "cron".to_string(),
+            trigger_detail: None,
+            started_at: Utc::now(),
+            completed_at: None,
+            status: RunStatus::Running,
+            result_summary: None,
+            tokens_used: None,
+            job_id: None,
+            created_at: Utc::now(),
+        };
+        db.create_routine_run(&run).await.expect("create run");
+
+        // Query for r1, r2 (doesn't exist), r3 (doesn't exist)
+        let counts = db
+            .count_running_routine_runs_batch(&[r1, r2, r3])
+            .await
+            .expect("batch query should work");
+
+        assert_eq!(counts.len(), 3, "Should have all 3 routine IDs");
+        assert_eq!(counts[&r1], 1, "r1 should have 1 running");
+        assert_eq!(counts[&r2], 0, "r2 should default to 0");
+        assert_eq!(counts[&r3], 0, "r3 should default to 0");
+    }
+
+    #[tokio::test]
+    async fn batch_query_only_counts_running_status() {
+        let (db, _tmp) = create_test_db().await;
+        let routine_id = Uuid::new_v4();
+
+        // Create routine
+        let routine = Routine {
+            id: routine_id,
+            name: "test-routine".to_string(),
+            description: "Test".to_string(),
+            user_id: "default".to_string(),
+            enabled: true,
+            trigger: Trigger::Cron {
+                schedule: "* * * * *".to_string(),
+                timezone: None,
+            },
+            action: RoutineAction::Lightweight {
+                prompt: "test".to_string(),
+                context_paths: vec![],
+                max_tokens: 1000,
+                use_tools: false,
+                max_tool_rounds: 3,
+            },
+            guardrails: RoutineGuardrails {
+                cooldown: std::time::Duration::from_secs(0),
+                max_concurrent: 5,
+                dedup_window: None,
+            },
+            notify: Default::default(),
+            last_run_at: None,
+            next_fire_at: None,
+            run_count: 0,
+            consecutive_failures: 0,
+            state: serde_json::json!({}),
+            created_at: Utc::now(),
+            updated_at: Utc::now(),
+        };
+        db.create_routine(&routine).await.expect("create routine");
+
+        // Create 5 runs with mixed statuses
+        let statuses = [
+            RunStatus::Running,
+            RunStatus::Running,
+            RunStatus::Ok,
+            RunStatus::Failed,
+            RunStatus::Attention,
+        ];
+
+        for status in statuses.iter() {
+            let run = RoutineRun {
+                id: Uuid::new_v4(),
+                routine_id,
+                trigger_type: "cron".to_string(),
+                trigger_detail: None,
+                started_at: Utc::now(),
+                completed_at: Some(Utc::now()),
+                status: *status,
+                result_summary: None,
+                tokens_used: None,
+                job_id: None,
+                created_at: Utc::now(),
+            };
+            db.create_routine_run(&run).await.expect("create run");
+        }
+
+        // Batch query should only count Running status
+        let counts = db
+            .count_running_routine_runs_batch(&[routine_id])
+            .await
+            .expect("batch query should work");
+
+        assert_eq!(
+            counts[&routine_id], 2,
+            "Should only count 2 Running status runs"
+        );
+    }
+
+    // -----------------------------------------------------------------------
+    // Test 2: Concurrent limit enforcement uses batch counts
+    // -----------------------------------------------------------------------
+
+    #[tokio::test]
+    async fn concurrent_limit_enforcement_with_batch_counts() {
+        let (db, _tmp) = create_test_db().await;
+
+        let r1 = Uuid::new_v4();
+        let r2 = Uuid::new_v4();
+
+        // Create 2 routines with max_concurrent=1 (r1) and max_concurrent=2 (r2)
+        for (routine_id, max_concurrent) in [(r1, 1), (r2, 2)] {
+            let routine = Routine {
+                id: routine_id,
+                name: format!("routine-{}", routine_id),
+                description: "Test".to_string(),
+                user_id: "default".to_string(),
+                enabled: true,
+                trigger: Trigger::Cron {
+                    schedule: "* * * * *".to_string(),
+                    timezone: None,
+                },
+                action: RoutineAction::Lightweight {
+                    prompt: "test".to_string(),
+                    context_paths: vec![],
+                    max_tokens: 1000,
+                    use_tools: false,
+                    max_tool_rounds: 3,
+                },
+                guardrails: RoutineGuardrails {
+                    cooldown: std::time::Duration::from_secs(0),
+                    max_concurrent,
+                    dedup_window: None,
+                },
+                notify: Default::default(),
+                last_run_at: None,
+                next_fire_at: None,
+                run_count: 0,
+                consecutive_failures: 0,
+                state: serde_json::json!({}),
+                created_at: Utc::now(),
+                updated_at: Utc::now(),
+            };
+            db.create_routine(&routine).await.expect("create routine");
+        }
+
+        // r1: create 1 running run (will hit max_concurrent=1)
+        let run = RoutineRun {
+            id: Uuid::new_v4(),
+            routine_id: r1,
+            trigger_type: "cron".to_string(),
+            trigger_detail: None,
+            started_at: Utc::now(),
+            completed_at: None,
+            status: RunStatus::Running,
+            result_summary: None,
+            tokens_used: None,
+            job_id: None,
+            created_at: Utc::now(),
+        };
+        db.create_routine_run(&run).await.expect("create run");
+
+        // r2: create 2 running runs (will hit max_concurrent=2)
+        for _ in 0..2 {
+            let run = RoutineRun {
+                id: Uuid::new_v4(),
+                routine_id: r2,
+                trigger_type: "cron".to_string(),
+                trigger_detail: None,
+                started_at: Utc::now(),
+                completed_at: None,
+                status: RunStatus::Running,
+                result_summary: None,
+                tokens_used: None,
+                job_id: None,
+                created_at: Utc::now(),
+            };
+            db.create_routine_run(&run).await.expect("create run");
+        }
+
+        // Batch query should return correct counts
+        let counts = db
+            .count_running_routine_runs_batch(&[r1, r2])
+            .await
+            .expect("batch query should work");
+
+        // Verify counts match the limits
+        assert_eq!(
+            counts[&r1], 1,
+            "r1 should have 1 running (at max_concurrent=1)"
+        );
+        assert_eq!(
+            counts[&r2], 2,
+            "r2 should have 2 running (at max_concurrent=2)"
+        );
+
+        // Now verify the limit enforcement logic
+        let r1_routine = db
+            .get_routine(r1)
+            .await
+            .expect("get routine")
+            .expect("routine exists");
+        let r2_routine = db
+            .get_routine(r2)
+            .await
+            .expect("get routine")
+            .expect("routine exists");
+
+        let r1_at_limit = counts[&r1] >= r1_routine.guardrails.max_concurrent as i64;
+        let r2_at_limit = counts[&r2] >= r2_routine.guardrails.max_concurrent as i64;
+
+        assert!(r1_at_limit, "r1 should be detected as at limit");
+        assert!(r2_at_limit, "r2 should be detected as at limit");
+
+        // If we add one more run to r2, it should exceed limit
+        let run = RoutineRun {
+            id: Uuid::new_v4(),
+            routine_id: r2,
+            trigger_type: "cron".to_string(),
+            trigger_detail: None,
+            started_at: Utc::now(),
+            completed_at: None,
+            status: RunStatus::Running,
+            result_summary: None,
+            tokens_used: None,
+            job_id: None,
+            created_at: Utc::now(),
+        };
+        db.create_routine_run(&run).await.expect("create run");
+
+        // Re-query to get updated counts
+        let counts = db
+            .count_running_routine_runs_batch(&[r1, r2])
+            .await
+            .expect("batch query should work");
+
+        let r2_exceeded_limit = counts[&r2] > r2_routine.guardrails.max_concurrent as i64;
+        assert!(r2_exceeded_limit, "r2 should have exceeded its limit");
+    }
+}

tests/e2e/scenarios/test_routine_event_batch.py

@@ -0,0 +1,534 @@
+"""
+E2E tests for event-triggered routines with batch loading.
+
+These tests verify that the N+1 query fix correctly:
+1. Fires event-triggered routines on matching messages
+2. Enforces concurrent limits via batch-loaded counts
+3. Maintains performance with multiple simultaneous triggers
+4. Works correctly through the full UI and agent loop
+
+Playwright-based UI tests + SSE verification.
+"""
+
+import asyncio
+import json
+import pytest
+from datetime import datetime, timedelta
+from typing import List, Dict, Any
+
+from playwright.async_api import async_playwright, Page, Browser, BrowserContext
+
+
+@pytest.fixture
+async def browser_and_context():
+    """Create a Playwright browser and context for testing."""
+    async with async_playwright() as p:
+        browser = await p.chromium.launch(headless=True)
+        context = await browser.new_context()
+        yield browser, context
+        await context.close()
+        await browser.close()
+
+
+class EventTriggerHelper:
+    """Helper methods for event trigger testing."""
+
+    def __init__(self, page: Page):
+        self.page = page
+
+    async def navigate_to_routines(self):
+        """Navigate to the routines page."""
+        await self.page.goto("http://localhost:8000/routines")
+        await self.page.wait_for_load_state("networkidle")
+
+    async def create_event_routine(
+        self,
+        name: str,
+        trigger_regex: str,
+        channel: str = "slack",
+        max_concurrent: int = 1,
+    ) -> str:
+        """
+        Create an event-triggered routine via UI.
+        Returns the routine ID.
+        """
+        await self.navigate_to_routines()
+
+        # Click "New Routine" button
+        await self.page.click('button:has-text("New Routine")')
+        await self.page.wait_for_selector('input[name="routine_name"]')
+
+        # Fill routine details
+        await self.page.fill('input[name="routine_name"]', name)
+        await self.page.fill(
+            'textarea[name="routine_description"]',
+            f"Test routine: {name}",
+        )
+
+        # Select "Event Trigger" type
+        await self.page.click('label:has-text("Event Trigger")')
+        await self.page.wait_for_selector('input[name="trigger_regex"]')
+
+        # Fill trigger details
+        await self.page.fill('input[name="trigger_regex"]', trigger_regex)
+        await self.page.select_option('select[name="trigger_channel"]', channel)
+
+        # Set guardrails
+        await self.page.fill('input[name="max_concurrent"]', str(max_concurrent))
+
+        # Select lightweight action
+        await self.page.click('label:has-text("Lightweight")')
+        await self.page.fill(
+            'textarea[name="lightweight_prompt"]',
+            "Acknowledge the message and confirm trigger worked.",
+        )
+
+        # Save routine
+        await self.page.click('button:has-text("Save Routine")')
+        await self.page.wait_for_selector('text=Routine created successfully')
+
+        # Extract routine ID from success message or URL
+        routine_id = await self.page.locator('data-testid=routine-id').text_content()
+        return routine_id.strip() if routine_id else None
+
+    async def create_multiple_routines(
+        self, base_name: str, count: int, trigger_regex: str = None
+    ) -> List[str]:
+        """Create multiple event-triggered routines."""
+        routine_ids = []
+        for i in range(count):
+            name = f"{base_name}_{i}"
+            regex = trigger_regex or f"({i}|{base_name})"
+            routine_id = await self.create_event_routine(name, regex)
+            routine_ids.append(routine_id)
+            await asyncio.sleep(0.1)  # Small delay between creations
+        return routine_ids
+
+    async def send_chat_message(self, message: str) -> List[str]:
+        """
+        Send a chat message and return SSE events received.
+        Captures all routine firing events.
+        """
+        await self.page.goto("http://localhost:8000/chat")
+        await self.page.wait_for_selector('input[placeholder*="message"]', timeout=5000)
+
+        # Collect SSE events
+        sse_events = []
+
+        async def capture_sse(response):
+            """Intercept SSE events."""
+            if "event-stream" in response.headers.get("content-type", ""):
+                text = await response.text()
+                for line in text.split("\n"):
+                    if line.startswith("data:"):
+                        try:
+                            event = json.loads(line[5:])
+                            sse_events.append(event)
+                        except json.JSONDecodeError:
+                            pass
+
+        self.page.on("response", capture_sse)
+
+        # Send message
+        await self.page.fill('input[placeholder*="message"]', message)
+        await self.page.press('input[placeholder*="message"]', "Enter")
+
+        # Wait for response
+        await self.page.wait_for_selector('text=Message processed', timeout=10000)
+        await asyncio.sleep(0.5)  # Allow time for SSE events
+
+        self.page.remove_listener("response", capture_sse)
+        return sse_events
+
+    async def get_routine_execution_log(self, routine_id: str) -> List[Dict]:
+        """Get execution log entries for a routine."""
+        await self.page.goto(f"http://localhost:8000/routines/{routine_id}/executions")
+        await self.page.wait_for_load_state("networkidle")
+
+        # Extract log entries from table
+        rows = await self.page.locator("tbody tr").all()
+        executions = []
+
+        for row in rows:
+            cells = await row.locator("td").all()
+            if len(cells) >= 3:
+                execution = {
+                    "timestamp": await cells[0].text_content(),
+                    "status": await cells[1].text_content(),
+                    "details": await cells[2].text_content(),
+                }
+                executions.append(execution)
+
+        return executions
+
+    async def check_database_queries_in_logs(
+        self, max_queries_expected: int = 1
+    ) -> int:
+        """Check debug logs for database query count."""
+        await self.page.goto("http://localhost:8000/debug/logs?filter=database")
+        await self.page.wait_for_load_state("networkidle")
+
+        # Count batch queries
+        log_lines = await self.page.locator("tr:has-text('batch')").all()
+        batch_count = len(log_lines)
+
+        # Count individual COUNT queries (should be 0 after fix)
+        count_queries = await self.page.locator("tr:has-text('COUNT')").all()
+        count_query_count = len(count_queries)
+
+        return batch_count, count_query_count
+
+
+# =============================================================================
+# Tests
+# =============================================================================
+
+
+@pytest.mark.asyncio
+async def test_create_event_trigger_routine(browser_and_context):
+    """Test creating an event-triggered routine via UI."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        routine_id = await helper.create_event_routine(
+            name="Test Trigger",
+            trigger_regex="test|demo",
+            channel="slack",
+            max_concurrent=1,
+        )
+
+        assert routine_id is not None, "Routine ID should be returned"
+        assert len(routine_id) > 0, "Routine ID should not be empty"
+
+    finally:
+        await page.close()
+
+
+@pytest.mark.asyncio
+async def test_event_trigger_fires_on_matching_message(browser_and_context):
+    """Test that event-triggered routine fires when message matches."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        # Create routine
+        routine_id = await helper.create_event_routine(
+            name="Alert Handler",
+            trigger_regex="urgent|critical|alert",
+            channel="slack",
+        )
+
+        # Send matching message
+        sse_events = await helper.send_chat_message("URGENT: Server down!")
+
+        # Verify routine fired (look for event in SSE stream)
+        routine_fired = any(
+            event.get("type") == "routine_fired" and event.get("routine_id") == routine_id
+            for event in sse_events
+        )
+        assert routine_fired, "Routine should fire on matching message"
+
+        # Check execution log
+        executions = await helper.get_routine_execution_log(routine_id)
+        assert len(executions) > 0, "Execution should be logged"
+        assert "success" in executions[0]["status"].lower()
+
+    finally:
+        await page.close()
+
+
+@pytest.mark.asyncio
+async def test_event_trigger_skips_non_matching_message(browser_and_context):
+    """Test that event-triggered routine skips when message doesn't match."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        # Create routine
+        routine_id = await helper.create_event_routine(
+            name="Alert Handler",
+            trigger_regex="urgent|critical|alert",
+            channel="slack",
+        )
+
+        # Send non-matching message
+        sse_events = await helper.send_chat_message("Hello, how are you?")
+
+        # Verify routine did NOT fire
+        routine_fired = any(
+            event.get("type") == "routine_fired" and event.get("routine_id") == routine_id
+            for event in sse_events
+        )
+        assert not routine_fired, "Routine should not fire on non-matching message"
+
+    finally:
+        await page.close()
+
+
+@pytest.mark.asyncio
+async def test_multiple_routines_fire_on_matching_message(browser_and_context):
+    """Test that multiple event-triggered routines fire on same message."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        # Create 3 overlapping routines
+        routine_ids = await helper.create_multiple_routines(
+            base_name="Handler", count=3, trigger_regex="alert|warning|error"
+        )
+
+        # Send matching message
+        sse_events = await helper.send_chat_message("ERROR: Database connection failed")
+
+        # Verify all 3 routines fired
+        fired_count = sum(
+            1
+            for event in sse_events
+            if event.get("type") == "routine_fired" and event.get("routine_id") in routine_ids
+        )
+
+        assert (
+            fired_count >= 3
+        ), f"Expected all 3 routines to fire, got {fired_count}"
+
+    finally:
+        await page.close()
+
+
+@pytest.mark.asyncio
+async def test_concurrent_limit_prevents_additional_fires(browser_and_context):
+    """Test that concurrent limit is enforced via batch counts."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        # Create routine with max_concurrent=1
+        routine_id = await helper.create_event_routine(
+            name="Limited Handler",
+            trigger_regex="process|task",
+            max_concurrent=1,
+        )
+
+        # Trigger first message
+        await helper.send_chat_message("Process message 1")
+        await asyncio.sleep(1)
+
+        # Check first execution logged
+        executions_1 = await helper.get_routine_execution_log(routine_id)
+        assert len(executions_1) >= 1
+
+        # Trigger second message while first is still running
+        sse_events = await helper.send_chat_message("Process message 2")
+
+        # Second routine should be skipped (concurrent limit)
+        routine_skipped = any(
+            event.get("type") == "routine_skipped"
+            and event.get("reason") == "max_concurrent_reached"
+            and event.get("routine_id") == routine_id
+            for event in sse_events
+        )
+        assert routine_skipped, "Routine should be skipped when concurrent limit reached"
+
+    finally:
+        await page.close()
+
+
+@pytest.mark.asyncio
+async def test_rapid_messages_with_multiple_triggers_efficiency(browser_and_context):
+    """Test efficiency of batch loading with multiple rapid messages."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        # Create 5 overlapping routines
+        routine_ids = await helper.create_multiple_routines(
+            base_name="Rapid", count=5, trigger_regex="test|demo|check"
+        )
+
+        # Send 10 matching messages rapidly
+        for i in range(10):
+            message = f"test message {i}"
+            await helper.send_chat_message(message)
+            await asyncio.sleep(0.1)
+
+        # Check database logs for query efficiency
+        batch_count, count_query_count = await helper.check_database_queries_in_logs()
+
+        # After fix: should have ~10 batch queries (1 per message)
+        # Before fix: would have ~50 individual COUNT queries (5 routines × 10 messages)
+        assert (
+            count_query_count == 0
+        ), f"Should have 0 individual COUNT queries after fix, got {count_query_count}"
+        assert (
+            batch_count <= 15
+        ), f"Should have <=15 batch queries for 10 messages, got {batch_count}"
+
+    finally:
+        await page.close()
+
+
+@pytest.mark.asyncio
+async def test_channel_filter_applied_correctly(browser_and_context):
+    """Test that channel filter prevents non-matching messages."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        # Create routine for Slack channel
+        slack_routine_id = await helper.create_event_routine(
+            name="Slack Handler",
+            trigger_regex="alert",
+            channel="slack",
+        )
+
+        # Simulate message from Telegram channel
+        # (Note: In real UI, would need to change channel context)
+        page.goto(
+            "http://localhost:8000/chat?channel=telegram"
+        )  # Switch channel
+        await helper.send_chat_message("alert: something urgent")
+
+        # Routine should not fire (different channel)
+        executions = await helper.get_routine_execution_log(slack_routine_id)
+
+        # Check if any recent execution (last 5 min) exists
+        recent = [
+            e
+            for e in executions
+            if (datetime.now() - datetime.fromisoformat(e["timestamp"])).total_seconds()
+            < 300
+        ]
+        assert (
+            len(recent) == 0
+        ), "Routine should not fire for different channel"
+
+    finally:
+        await page.close()
+
+
+@pytest.mark.asyncio
+async def test_batch_query_failure_handling(browser_and_context):
+    """Test graceful handling of batch query failures."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        # Create routine
+        routine_id = await helper.create_event_routine(
+            name="Error Handler",
+            trigger_regex="test",
+        )
+
+        # Simulate database error in logs (if possible with test hooks)
+        # For now, just verify error handling doesn't crash UI
+        await helper.send_chat_message("test message")
+
+        # Check that UI remains responsive
+        assert await page.locator("text=Message processed").is_visible()
+
+    finally:
+        await page.close()
+
+
+@pytest.mark.asyncio
+async def test_routine_execution_history_display(browser_and_context):
+    """Test that execution history correctly displays routine firings."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        # Create routine
+        routine_id = await helper.create_event_routine(
+            name="History Test",
+            trigger_regex="test",
+        )
+
+        # Trigger routine 3 times
+        for i in range(3):
+            await helper.send_chat_message(f"test message {i}")
+            await asyncio.sleep(0.2)
+
+        # Check execution log
+        executions = await helper.get_routine_execution_log(routine_id)
+        assert len(executions) >= 3, "Should have at least 3 executions logged"
+
+        # Verify all are recent (within last 5 minutes)
+        for execution in executions[:3]:
+            timestamp = datetime.fromisoformat(execution["timestamp"])
+            age = datetime.now() - timestamp
+            assert age < timedelta(minutes=5), "Execution should be recent"
+
+    finally:
+        await page.close()
+
+
+@pytest.mark.asyncio
+async def test_concurrent_batch_loads_independent(browser_and_context):
+    """Test that concurrent messages each get independent batch queries."""
+    browser, context = browser_and_context
+    page = await context.new_page()
+    helper = EventTriggerHelper(page)
+
+    try:
+        # Create 5 routines matching different patterns
+        r1_id = await helper.create_event_routine(
+            name="Pattern A", trigger_regex="alpha|alpha_only"
+        )
+        r2_id = await helper.create_event_routine(
+            name="Pattern B", trigger_regex="beta|beta_only"
+        )
+        r3_id = await helper.create_event_routine(
+            name="Pattern AB", trigger_regex="alpha|beta|common"
+        )
+
+        # Send overlapping messages
+        # Message 1: matches r1, r3
+        sse1 = await helper.send_chat_message("alpha common")
+        await asyncio.sleep(0.1)
+
+        # Message 2: matches r2, r3
+        sse2 = await helper.send_chat_message("beta common")
+        await asyncio.sleep(0.1)
+
+        # Verify correct routines fired
+        r1_fired_msg1 = any(
+            e.get("routine_id") == r1_id for e in sse1 if e.get("type") == "routine_fired"
+        )
+        r2_fired_msg2 = any(
+            e.get("routine_id") == r2_id for e in sse2 if e.get("type") == "routine_fired"
+        )
+        r3_fired_both = (
+            any(
+                e.get("routine_id") == r3_id for e in sse1 if e.get("type") == "routine_fired"
+            )
+            and any(
+                e.get("routine_id") == r3_id for e in sse2 if e.get("type") == "routine_fired"
+            )
+        )
+
+        assert r1_fired_msg1, "Routine 1 should fire on message 1"
+        assert r2_fired_msg2, "Routine 2 should fire on message 2"
+        assert r3_fired_both, "Routine 3 should fire on both messages"
+
+    finally:
+        await page.close()
+
+
+# =============================================================================
+# Integration with existing test patterns
+# =============================================================================
+
+
+if __name__ == "__main__":
+    # Run tests with: pytest tests/e2e/scenarios/test_routine_event_batch.py -v
+    pytest.main([__file__, "-v", "-s"])