scheduler: use per-group average duration for unknown tests

.beads/issues.jsonl

@@ -15,6 +15,7 @@
 {"id":"code-110","title":"Create skills/offload/SKILL.md","description":"Create the /offload skill SKILL.md with frontmatter, overview, invocation guide, decision guide, exit codes, debugging, CLI reference, and config groups reference (~150 lines)","status":"done","priority":2,"issue_type":"task","created_at":"2026-03-12T10:58:56.785986-07:00","created_by":"danver","updated_at":"2026-03-12T11:03:25.317108-07:00"}
 {"id":"code-111","title":"Create install-skills.sh","description":"Create standalone bash installer that works both curl|bash and local. Symlinks in-repo, downloads from GitHub raw URLs standalone. Installs both /offload and /offload-onboard skills.","status":"done","priority":2,"issue_type":"task","created_at":"2026-03-12T10:59:01.642817-07:00","created_by":"danver","updated_at":"2026-03-12T11:09:18.367138-07:00"}
 {"id":"code-112","title":"Modify justfile to delegate to install-skills.sh","description":"Replace hardcoded install-skill recipe (lines 39-55) with install-skills recipe that calls ./install-skills.sh, plus install-skill alias for backward compat.","status":"done","priority":2,"issue_type":"task","created_at":"2026-03-12T10:59:05.717901-07:00","created_by":"danver","updated_at":"2026-03-12T11:11:51.621362-07:00"}
+{"id":"code-104b","title":"Replace flat default_duration with per-group average in scheduler","description":"Replace the hardcoded 1s default_duration in schedule_lpt with per-group average durations computed from historical data. Steps: (1) Add group() accessor to TestInstance. (2) Change schedule_lpt signature to accept HashMap\u003cString, Duration\u003e for group defaults instead of a single Duration. (3) In orchestrator.rs, compute per-group averages from the durations map + test records, and pass that mapping to schedule_lpt. (4) Update all scheduler tests. Keep the 1s fallback only when a group has zero historical data.","status":"done","priority":1,"issue_type":"task","owner":"jacob.kirmayer@imbue.com","created_at":"2026-03-17T12:02:13.11179-07:00","created_by":"Jacob Kirmayer","updated_at":"2026-03-17T12:07:04.910208-07:00"}
 {"id":"code-11","title":"Rename project: Rename offload-*.toml config files to offload-*.toml","description":"Rename all configuration files with 'offload' prefix to use 'offload' prefix:\n- offload.toml -\u003e offload.toml\n- offload-local.toml -\u003e offload-local.toml\n- offload-modal.toml -\u003e offload-modal.toml\n- offload-cargo-local.toml -\u003e offload-cargo-local.toml\n- offload-cargo-modal.toml -\u003e offload-cargo-modal.toml\n- offload-computronium-modal.toml -\u003e offload-computronium-modal.toml\n- offload-sculptor-modal.toml -\u003e offload-sculptor-modal.toml\n\nAlso update the [offload] section in these files to [offload].","status":"done","priority":1,"issue_type":"task","created_at":"2026-01-29T18:25:03.560121502Z","created_by":"Danver Braganza","updated_at":"2026-01-29T18:45:18.15783543Z"}
 {"id":"code-12","title":"Rename project: Update README.md from offload to offload","description":"Update README.md to replace all references to 'offload' with 'offload'. This includes:\n- Project title\n- Feature descriptions\n- Installation commands\n- CLI examples (offload init, offload run, etc.)\n- Configuration file references (offload.toml -\u003e offload.toml)\n- Example configuration sections ([offload] -\u003e [offload])\n- All documentation text","status":"done","priority":1,"issue_type":"task","created_at":"2026-01-29T18:25:08.706866046Z","created_by":"Danver Braganza","updated_at":"2026-01-29T18:50:11.476117046Z"}
 {"id":"code-13","title":"Rename project: Update scripts/modal_sandbox.py from offload to offload","description":"Update scripts/modal_sandbox.py to replace all references to 'offload' with 'offload'. This includes:\n- Module docstring\n- CLI help text\n- Modal App names (offload-sandbox -\u003e offload-sandbox, offload-rust-sandbox -\u003e offload-rust-sandbox, etc.)\n- Function docstrings\n- Comments","status":"done","priority":1,"issue_type":"task","created_at":"2026-01-29T18:25:14.017333924Z","created_by":"Danver Braganza","updated_at":"2026-01-29T18:52:06.241321461Z"}

src/framework.rs

@@ -121,6 +121,10 @@ impl<'a> TestInstance<'a> {
     pub fn id(&self) -> &str {
         &self.record.id
     }
+
+    pub fn group(&self) -> &str {
+        &self.record.group
+    }
 }
 
 /// The result of executing a single test.

src/orchestrator.rs

@@ -4,7 +4,7 @@ pub mod runner;
 pub mod scheduler;
 pub mod spawn;
 
-use std::collections::VecDeque;
+use std::collections::{HashMap, VecDeque};
 use std::sync::Arc;
 use std::sync::atomic::AtomicBool;
 use std::time::Duration;
@@ -257,11 +257,27 @@ where
                 junit_path.display()
             );
         }
-        // Default duration for unknown tests: 1 second (conservative estimate)
+        // Compute per-group average durations for tests without historical data
+        let group_to_default_duration = {
+            let mut group_totals: HashMap<String, (Duration, usize)> = HashMap::new();
+            for test in &tests_to_run {
+                if let Some(&d) = durations.get(test.id()) {
+                    let entry = group_totals
+                        .entry(test.group().to_string())
+                        .or_insert((Duration::ZERO, 0));
+                    entry.0 += d;
+                    entry.1 += 1;
+                }
+            }
+            group_totals
+                .into_iter()
+                .map(|(group, (total, count))| (group, total / count as u32))
+                .collect::<HashMap<String, Duration>>()
+        };
         let batches = scheduler.schedule(
             &tests_to_run,
             &durations,
-            Duration::from_secs(1),
+            &group_to_default_duration,
             Some(MAX_BATCH_DURATION),
         );
         drop(_sched_span);

src/orchestrator/scheduler.rs

@@ -100,8 +100,9 @@ impl Scheduler {
     ///
     /// * `tests` - Tests to schedule
     /// * `durations` - Historical test durations from previous runs.
-    ///   Tests not in the map use `default_duration`.
-    /// * `default_duration` - Duration to use for tests without historical data.
+    ///   Tests not in the map use the per-group average from `group_to_default_duration`.
+    /// * `group_to_default_duration` - Per-group average duration for tests without historical data.
+    ///   Falls back to 1 second if the group has no entry.
     /// * `max_batch_duration` - Optional cap on the total duration of each batch.
     ///   A single test that exceeds the cap is still placed alone in its own batch.
     ///
@@ -120,7 +121,7 @@ impl Scheduler {
         &self,
         tests: &[TestInstance<'a>],
         durations: &HashMap<String, Duration>,
-        default_duration: Duration,
+        group_to_default_duration: &HashMap<String, Duration>,
         max_batch_duration: Option<Duration>,
     ) -> Vec<Vec<TestInstance<'a>>> {
         if tests.is_empty() {
@@ -134,11 +135,17 @@ impl Scheduler {
                 let duration = match durations.get(t.id()) {
                     Some(&d) => d,
                     None => {
+                        let fallback = group_to_default_duration
+                            .get(t.group())
+                            .copied()
+                            .unwrap_or(Duration::from_secs(1));
                         tracing::warn!(
-                            "No historical duration for test '{}', using default",
-                            t.id()
+                            "No historical duration for test '{}', using group '{}' default {:?}",
+                            t.id(),
+                            t.group(),
+                            fallback,
                         );
-                        default_duration
+                        fallback
                     }
                 };
                 (*t, duration)
@@ -192,7 +199,7 @@ mod tests {
         let scheduler = Scheduler::new(4);
         let durations = HashMap::new();
         let batches: Vec<Vec<TestInstance>> =
-            scheduler.schedule(&[], &durations, Duration::from_secs(1), None);
+            scheduler.schedule(&[], &durations, &HashMap::new(), None);
         assert!(batches.is_empty());
     }
 
@@ -211,7 +218,7 @@ mod tests {
         durations.insert("medium_test".to_string(), Duration::from_secs(5));
         durations.insert("fast_test".to_string(), Duration::from_secs(1));
 
-        let batches = scheduler.schedule(&tests, &durations, Duration::from_secs(1), None);
+        let batches = scheduler.schedule(&tests, &durations, &HashMap::new(), None);
 
         // With LPT:
         // 1. Assign slow_test (10s) to worker 0 -> loads: [10, 0]
@@ -241,7 +248,7 @@ mod tests {
         durations.insert("test_b".to_string(), Duration::from_secs(5));
         durations.insert("test_c".to_string(), Duration::from_secs(3));
 
-        let batches = scheduler.schedule(&tests, &durations, Duration::from_secs(1), None);
+        let batches = scheduler.schedule(&tests, &durations, &HashMap::new(), None);
 
         // Each test in its own batch (3 workers, 3 tests)
         // Sorted by duration: test_b (5s), test_c (3s), test_a (1s)
@@ -264,7 +271,7 @@ mod tests {
         durations.insert("known_slow".to_string(), Duration::from_secs(10));
         // unknown_test will use default of 1 second
 
-        let batches = scheduler.schedule(&tests, &durations, Duration::from_secs(1), None);
+        let batches = scheduler.schedule(&tests, &durations, &HashMap::new(), None);
 
         assert_eq!(batches.len(), 2);
         // known_slow (10s) should be in heaviest batch
@@ -286,7 +293,7 @@ mod tests {
         let mut durations = HashMap::new();
         durations.insert("test_a".to_string(), Duration::from_secs(5));
 
-        let batches = scheduler.schedule(&tests, &durations, Duration::from_secs(1), None);
+        let batches = scheduler.schedule(&tests, &durations, &HashMap::new(), None);
 
         // Each instance of test_a must be in a different batch
         assert_eq!(batches.len(), 3);
@@ -319,7 +326,7 @@ mod tests {
         durations.insert("test_b".to_string(), Duration::from_secs(5));
         durations.insert("test_c".to_string(), Duration::from_secs(1));
 
-        let batches = scheduler.schedule(&tests, &durations, Duration::from_secs(1), None);
+        let batches = scheduler.schedule(&tests, &durations, &HashMap::new(), None);
 
         // Verify no batch contains duplicate test IDs
         for batch in &batches {
@@ -348,7 +355,7 @@ mod tests {
         let tests: Vec<_> = records.iter().map(|r| r.test()).collect();
 
         let durations = HashMap::new();
-        let batches = scheduler.schedule(&tests, &durations, Duration::from_secs(1), None);
+        let batches = scheduler.schedule(&tests, &durations, &HashMap::new(), None);
 
         // Each instance must be in a separate batch
         assert_eq!(batches.len(), 3);
@@ -379,7 +386,7 @@ mod tests {
         let batches = scheduler.schedule(
             &tests,
             &durations,
-            Duration::from_secs(1),
+            &HashMap::new(),
             Some(MAX_BATCH_DURATION),
         );
 
@@ -418,7 +425,7 @@ mod tests {
         let batches = scheduler.schedule(
             &tests,
             &durations,
-            Duration::from_secs(1),
+            &HashMap::new(),
             Some(MAX_BATCH_DURATION),
         );
 
@@ -453,7 +460,7 @@ mod tests {
         let batches = scheduler.schedule(
             &tests,
             &durations,
-            Duration::from_secs(1),
+            &HashMap::new(),
             Some(MAX_BATCH_DURATION),
         );
 
@@ -489,7 +496,7 @@ mod tests {
         ];
         let tests: Vec<_> = records.iter().map(|r| r.test()).collect();
 
-        let batches = scheduler.schedule(&tests, &HashMap::new(), Duration::from_secs(1), None);
+        let batches = scheduler.schedule(&tests, &HashMap::new(), &HashMap::new(), None);
 
         // Two tests that each use >half the command length budget must be in separate batches
         assert_eq!(batches.len(), 2);
@@ -506,7 +513,7 @@ mod tests {
             .collect();
         let tests: Vec<_> = records.iter().map(|r| r.test()).collect();
 
-        let batches = scheduler.schedule(&tests, &HashMap::new(), Duration::from_secs(0), None);
+        let batches = scheduler.schedule(&tests, &HashMap::new(), &HashMap::new(), None);
 
         // Total command length is ~400 chars, well under 30k — should be 1 batch
         assert_eq!(batches.len(), 1);