fix(flue): Fix span nesting (#2098)

lforst · web-flow · commit e563ea78c36f · 2026-06-15T10:56:42.000+02:00
- Nests spans created inside workflow under workflow spans - Nests spans created inside tools under tool span <img width="273" height="415" alt="Screenshot 2026-06-12 at 11 23 09" src="https://github.com/user-attachments/assets/88275889-6746-4fa6-bd2c-30aace861e34" />
diff --git a/.changeset/light-days-brush.md b/.changeset/light-days-brush.md
@@ -0,0 +1,5 @@
+---
+"braintrust": patch
+---
+
+fix(flue): Fix span nesting
diff --git a/e2e/scenarios/flue-instrumentation/.flue/workflows/instrumentation.mjs b/e2e/scenarios/flue-instrumentation/.flue/workflows/instrumentation.mjs
@@ -1,5 +1,6 @@
 import { createAgent, Type } from "@flue/runtime";
 import { local } from "@flue/runtime/node";
+import { traced } from "braintrust";
 import {
   FLUE_MODEL,
   FLUE_REASONING_MODEL,
@@ -48,12 +49,27 @@ const flueE2EAgent = createAgent(() => ({
 const lookupTool = {
   description:
     "Return a deterministic lookup result with an id needed by web_search.",
-  execute: async (args) =>
-    JSON.stringify({
+  execute: async (args) => {
+    await traced(
+      async (span) => {
+        span.log({ output: "lookup-active" });
+      },
+      {
+        name: "flue.toolCurrentProbe",
+        event: {
+          metadata: {
+            scenario: SCENARIO_NAME,
+          },
+        },
+      },
+    );
+
+    return JSON.stringify({
       id: "flue-session-2026",
       query: args.query,
       topic: "session instrumentation",
-    }),
+    });
+  },
   name: "lookup",
   parameters: Type.Object({
     query: Type.String(),
@@ -101,6 +117,20 @@ export async function route(_ctx, next) {
 }
 
 export async function run({ init, payload }) {
+  await traced(
+    async (span) => {
+      span.log({ output: "active" });
+    },
+    {
+      name: "flue.workflowCurrentProbe",
+      event: {
+        metadata: {
+          scenario: SCENARIO_NAME,
+        },
+      },
+    },
+  );
+
   const harness = await init(flueE2EAgent, { name: "default" });
   const session = await harness.session("main");
   const skillSession = await harness.session("skill");
diff --git a/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-auto-hook.span-tree.json b/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-auto-hook.span-tree.json
@@ -4,6 +4,14 @@
       "name": "workflow:instrumentation",
       "type": "task",
       "children": [
+        {
+          "name": "flue.workflowCurrentProbe",
+          "children": [],
+          "output": "active",
+          "metadata": {
+            "scenario": "flue-instrumentation"
+          }
+        },
         {
           "name": "flue.prompt",
           "type": "task",
@@ -68,7 +76,16 @@
             {
               "name": "tool:lookup",
               "type": "tool",
-              "children": [],
+              "children": [
+                {
+                  "name": "flue.toolCurrentProbe",
+                  "children": [],
+                  "output": "lookup-active",
+                  "metadata": {
+                    "scenario": "flue-instrumentation"
+                  }
+                }
+              ],
               "input": {
                 "query": "flue instrumentation"
               },
diff --git a/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-auto-hook.span-tree.txt b/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-auto-hook.span-tree.txt
@@ -19,6 +19,11 @@ span_tree:
     metrics: {
       "duration_ms": 0
     }
+    ├── flue.workflowCurrentProbe
+    │   output: "active"
+    │   metadata: {
+    │     "scenario": "flue-instrumentation"
+    │   }
     ├── flue.prompt [task]
     │   input: [
     │     {
@@ -116,6 +121,11 @@ span_tree:
     │   │   metrics: {
     │   │     "duration_ms": 0
     │   │   }
+    │   │   └── flue.toolCurrentProbe
+    │   │       output: "lookup-active"
+    │   │       metadata: {
+    │   │         "scenario": "flue-instrumentation"
+    │   │       }
     │   ├── llm:claude-sonnet-4-5-20250929 [llm]
     │   │   input: [
     │   │     {
diff --git a/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-cli.span-tree.json b/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-cli.span-tree.json
@@ -4,6 +4,14 @@
       "name": "workflow:instrumentation",
       "type": "task",
       "children": [
+        {
+          "name": "flue.workflowCurrentProbe",
+          "children": [],
+          "output": "active",
+          "metadata": {
+            "scenario": "flue-instrumentation"
+          }
+        },
         {
           "name": "flue.prompt",
           "type": "task",
@@ -68,7 +76,16 @@
             {
               "name": "tool:lookup",
               "type": "tool",
-              "children": [],
+              "children": [
+                {
+                  "name": "flue.toolCurrentProbe",
+                  "children": [],
+                  "output": "lookup-active",
+                  "metadata": {
+                    "scenario": "flue-instrumentation"
+                  }
+                }
+              ],
               "input": {
                 "query": "flue instrumentation"
               },
diff --git a/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-cli.span-tree.txt b/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-cli.span-tree.txt
@@ -19,6 +19,11 @@ span_tree:
     metrics: {
       "duration_ms": 0
     }
+    ├── flue.workflowCurrentProbe
+    │   output: "active"
+    │   metadata: {
+    │     "scenario": "flue-instrumentation"
+    │   }
     ├── flue.prompt [task]
     │   input: [
     │     {
@@ -116,6 +121,11 @@ span_tree:
     │   │   metrics: {
     │   │     "duration_ms": 0
     │   │   }
+    │   │   └── flue.toolCurrentProbe
+    │   │       output: "lookup-active"
+    │   │       metadata: {
+    │   │         "scenario": "flue-instrumentation"
+    │   │       }
     │   ├── llm:claude-sonnet-4-5-20250929 [llm]
     │   │   input: [
     │   │     {
diff --git a/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-explicit.span-tree.json b/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-explicit.span-tree.json
@@ -4,6 +4,14 @@
       "name": "workflow:instrumentation",
       "type": "task",
       "children": [
+        {
+          "name": "flue.workflowCurrentProbe",
+          "children": [],
+          "output": "active",
+          "metadata": {
+            "scenario": "flue-instrumentation"
+          }
+        },
         {
           "name": "flue.prompt",
           "type": "task",
@@ -68,7 +76,16 @@
             {
               "name": "tool:lookup",
               "type": "tool",
-              "children": [],
+              "children": [
+                {
+                  "name": "flue.toolCurrentProbe",
+                  "children": [],
+                  "output": "lookup-active",
+                  "metadata": {
+                    "scenario": "flue-instrumentation"
+                  }
+                }
+              ],
               "input": {
                 "query": "flue instrumentation"
               },
diff --git a/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-explicit.span-tree.txt b/e2e/scenarios/flue-instrumentation/__snapshots__/flue-v0-8-0-explicit.span-tree.txt
@@ -19,6 +19,11 @@ span_tree:
     metrics: {
       "duration_ms": 0
     }
+    ├── flue.workflowCurrentProbe
+    │   output: "active"
+    │   metadata: {
+    │     "scenario": "flue-instrumentation"
+    │   }
     ├── flue.prompt [task]
     │   input: [
     │     {
@@ -116,6 +121,11 @@ span_tree:
     │   │   metrics: {
     │   │     "duration_ms": 0
     │   │   }
+    │   │   └── flue.toolCurrentProbe
+    │   │       output: "lookup-active"
+    │   │       metadata: {
+    │   │         "scenario": "flue-instrumentation"
+    │   │       }
     │   ├── llm:claude-sonnet-4-5-20250929 [llm]
     │   │   input: [
     │   │     {
diff --git a/e2e/scenarios/flue-instrumentation/assertions.ts b/e2e/scenarios/flue-instrumentation/assertions.ts
@@ -160,19 +160,26 @@ function isFlueChildSpan(event: CapturedLogEvent): boolean {
     event.span.name?.startsWith("tool:") === true ||
     event.span.name?.startsWith("task:") === true ||
     event.span.name?.startsWith("compaction:") === true ||
+    event.span.name === "flue.toolCurrentProbe" ||
     event.span.name === "flue.task"
   );
 }
 
 function buildSpanTree(events: CapturedLogEvent[]): SpanTreeEntry[] {
   const workflow = findLatestSpanByPrefix(events, "workflow:");
+  const workflowCurrentProbe = findLatestChildSpan(
+    events,
+    "flue.workflowCurrentProbe",
+    workflow?.span.id,
+  );
   const promptSpan = findFlueOperation(events, "flue.prompt");
   const skillSpan = findFlueOperation(events, "flue.skill");
   const taskSpan = findFlueOperation(events, "flue.task");
   const compactSpan = findFlueOperation(events, "flue.compact");
 
   return [
     workflow,
+    workflowCurrentProbe,
     promptSpan,
     ...findFlueDescendants(events, promptSpan, isFlueChildSpan),
     skillSpan,
@@ -223,6 +230,41 @@ export function defineFlueInstrumentationAssertions(options: {
       });
     });
 
+    test(
+      "makes the Flue workflow span current for app spans",
+      testConfig,
+      () => {
+        const workflow = findLatestSpanByPrefix(events, "workflow:");
+        const appSpan = findLatestChildSpan(
+          events,
+          "flue.workflowCurrentProbe",
+          workflow?.span.id,
+        );
+
+        expect(appSpan).toBeDefined();
+        expect(appSpan?.span.parentIds).toEqual([workflow?.span.id]);
+        expect(appSpan?.output).toBe("active");
+      },
+    );
+
+    test("makes Flue tool spans current for app spans", testConfig, () => {
+      const promptSpan = findFlueOperation(events, "flue.prompt");
+      const lookupToolSpan = findFlueDescendants(
+        events,
+        promptSpan,
+        (event) => event.span.name === "tool:lookup",
+      )[0];
+      const appSpan = findLatestChildSpan(
+        events,
+        "flue.toolCurrentProbe",
+        lookupToolSpan?.span.id,
+      );
+
+      expect(appSpan).toBeDefined();
+      expect(appSpan?.span.parentIds).toEqual([lookupToolSpan?.span.id]);
+      expect(appSpan?.output).toBe("lookup-active");
+    });
+
     test("captures observe-based Flue operation spans", testConfig, () => {
       for (const flueSpanName of [
         "flue.prompt",
diff --git a/js/src/instrumentation/plugins/flue-plugin.test.ts b/js/src/instrumentation/plugins/flue-plugin.test.ts
diff --git a/js/src/instrumentation/plugins/flue-plugin.ts b/js/src/instrumentation/plugins/flue-plugin.ts

-Original file line number
+Diff line change
@@ @@ -0,0 +1,5 @@ @@
 +---
 +"braintrust": patch
 +---
++
 +fix(flue): Fix span nesting