fix(memory,core): eliminate Qdrant upsert timing race and add RuntimeLayer tests (#2413, #2361) (#2430)

- use wait=true on QdrantOps::upsert to block until points are indexed,
  eliminating scroll_all race in document_integration tests (#2413)
- add multi_layer_before_after_tool_ordering test to RuntimeLayer (#2361)
- fix float comparison in confusability threshold test (clippy::float_cmp)

Author: bug-ops

CHANGELOG.md

@@ -19,6 +19,12 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
 ### Fixed
 
+- fix(memory): use `wait=true` on Qdrant upsert to eliminate testcontainer timing race — points are now indexed and queryable immediately after `upsert` returns (closes #2413)
+
+### Added (tests)
+
+- test(core): add `multi_layer_before_after_tool_ordering` — verifies both layers are called in FIFO order for `before_tool` and `after_tool` (#2361)
+
 - fix(classifiers): sha2 0.11 hex formatting — replace `format!("{:x}", ...)` with `hex::encode(...)` in `verify_sha256` and its test helper (#2401)
 - fix(deps): bump sha2 0.10→0.11, ordered-float 5.1→5.3, proptest 1.10→1.11, toml 1.0→1.1, uuid 1.22→1.23 (#2401)
 - fix(skills): `two_stage_matching` and `confusability_threshold` config fields are now applied at agent startup; `AgentBuilder` gains `with_two_stage_matching` and `with_confusability_threshold` builder methods wired in `runner.rs` and `daemon.rs` (closes #2404)

crates/zeph-core/src/agent/builder.rs

@@ -1669,14 +1669,14 @@ mod tests {
         );
 
         let agent = make_agent().with_confusability_threshold(1.5);
-        assert_eq!(
-            agent.skill_state.confusability_threshold, 1.0,
+        assert!(
+            (agent.skill_state.confusability_threshold - 1.0).abs() < f32::EPSILON,
             "with_confusability_threshold must clamp values above 1.0"
         );
 
         let agent = make_agent().with_confusability_threshold(-0.1);
-        assert_eq!(
-            agent.skill_state.confusability_threshold, 0.0,
+        assert!(
+            agent.skill_state.confusability_threshold.abs() < f32::EPSILON,
             "with_confusability_threshold must clamp values below 0.0"
         );
     }

crates/zeph-core/src/runtime_layer.rs

@@ -344,6 +344,81 @@ mod tests {
         );
     }
 
+    /// Two layers registered in order [A, B]: `before_tool` must fire A then B,
+    /// and `after_tool` must fire A then B (forward order for both).
+    #[tokio::test]
+    async fn multi_layer_before_after_tool_ordering() {
+        use std::sync::{Arc, Mutex};
+
+        struct ToolOrderLayer {
+            id: u32,
+            log: Arc<Mutex<Vec<String>>>,
+        }
+        impl RuntimeLayer for ToolOrderLayer {
+            fn before_tool<'a>(
+                &'a self,
+                _ctx: &'a LayerContext<'_>,
+                _call: &'a ToolCall,
+            ) -> Pin<Box<dyn Future<Output = BeforeToolResult> + Send + 'a>> {
+                self.log
+                    .lock()
+                    .unwrap()
+                    .push(format!("before_tool_{}", self.id));
+                Box::pin(std::future::ready(None))
+            }
+
+            fn after_tool<'a>(
+                &'a self,
+                _ctx: &'a LayerContext<'_>,
+                _call: &'a ToolCall,
+                _result: &'a Result<Option<ToolOutput>, ToolError>,
+            ) -> Pin<Box<dyn Future<Output = ()> + Send + 'a>> {
+                self.log
+                    .lock()
+                    .unwrap()
+                    .push(format!("after_tool_{}", self.id));
+                Box::pin(std::future::ready(()))
+            }
+        }
+
+        let log: Arc<Mutex<Vec<String>>> = Arc::new(Mutex::new(Vec::new()));
+        let layer_a = ToolOrderLayer {
+            id: 1,
+            log: Arc::clone(&log),
+        };
+        let layer_b = ToolOrderLayer {
+            id: 2,
+            log: Arc::clone(&log),
+        };
+
+        let ctx = LayerContext {
+            conversation_id: None,
+            turn_number: 0,
+        };
+        let call = ToolCall {
+            tool_id: "shell".into(),
+            params: serde_json::Map::new(),
+        };
+        let result: Result<Option<ToolOutput>, ToolError> = Ok(None);
+
+        layer_a.before_tool(&ctx, &call).await;
+        layer_b.before_tool(&ctx, &call).await;
+        layer_a.after_tool(&ctx, &call, &result).await;
+        layer_b.after_tool(&ctx, &call, &result).await;
+
+        let events = log.lock().unwrap().clone();
+        assert_eq!(
+            events,
+            vec![
+                "before_tool_1",
+                "before_tool_2",
+                "after_tool_1",
+                "after_tool_2"
+            ],
+            "tool hooks must fire in registration order"
+        );
+    }
+
     /// `NoopLayer` `after_tool` returns `()` without errors.
     #[tokio::test]
     async fn noop_layer_after_tool_returns_unit() {

crates/zeph-memory/src/qdrant_ops.rs

@@ -143,10 +143,8 @@ impl QdrantOps {
     ///
     /// Returns an error if the upsert fails.
     pub async fn upsert(&self, collection: &str, points: Vec<PointStruct>) -> QdrantResult<()> {
-        // wait(false): fire-and-forget on the hot path — saves 3-15ms per embedding store call.
-        // Qdrant guarantees eventual consistency; read-your-writes is not required here.
         self.client
-            .upsert_points(UpsertPointsBuilder::new(collection, points).wait(false))
+            .upsert_points(UpsertPointsBuilder::new(collection, points).wait(true))
             .await
             .map_err(Box::new)?;
         Ok(())