fix: use backend thread token usage for header total (#2800)

* fix: use backend thread token usage for header total

* Refactor thread token usage fetch

backend/tests/test_thread_token_usage.py

@@ -0,0 +1,55 @@
+"""Tests for thread-level token usage aggregation API."""
+
+from __future__ import annotations
+
+from unittest.mock import AsyncMock, MagicMock
+
+from _router_auth_helpers import make_authed_test_app
+from fastapi.testclient import TestClient
+
+from app.gateway.routers import thread_runs
+
+
+def _make_app(run_store: MagicMock):
+    app = make_authed_test_app()
+    app.include_router(thread_runs.router)
+    app.state.run_store = run_store
+    return app
+
+
+def test_thread_token_usage_returns_stable_shape():
+    run_store = MagicMock()
+    run_store.aggregate_tokens_by_thread = AsyncMock(
+        return_value={
+            "total_tokens": 150,
+            "total_input_tokens": 90,
+            "total_output_tokens": 60,
+            "total_runs": 2,
+            "by_model": {"unknown": {"tokens": 150, "runs": 2}},
+            "by_caller": {
+                "lead_agent": 120,
+                "subagent": 25,
+                "middleware": 5,
+            },
+        },
+    )
+    app = _make_app(run_store)
+
+    with TestClient(app) as client:
+        response = client.get("/api/threads/thread-1/token-usage")
+
+    assert response.status_code == 200
+    assert response.json() == {
+        "thread_id": "thread-1",
+        "total_tokens": 150,
+        "total_input_tokens": 90,
+        "total_output_tokens": 60,
+        "total_runs": 2,
+        "by_model": {"unknown": {"tokens": 150, "runs": 2}},
+        "by_caller": {
+            "lead_agent": 120,
+            "subagent": 25,
+            "middleware": 5,
+        },
+    }
+    run_store.aggregate_tokens_by_thread.assert_awaited_once_with("thread-1")