fix: ensure LLM stats tracking is accurate by including completed subagents (#441 )

Add Strix GitHub Actions integration tip
2026-04-13 00:09:13 -04:00 · 2026-04-12 12:43:41 -07:00
7 changed files with 160 additions and 42 deletions
--- a/README.md
+++ b/README.md
@@ -32,9 +32,8 @@
 </div>


-
 > [!TIP]
-> **New!** Strix integrates seamlessly with GitHub Actions and CI/CD pipelines. Automatically scan for vulnerabilities on every pull request and block insecure code before it reaches production!
+> **New!** Strix integrates seamlessly with GitHub Actions and CI/CD pipelines. Automatically scan for vulnerabilities on every pull request and block insecure code before it reaches production - [Get started with no setup required](https://app.strix.ai).

 ---

--- a/strix/agents/base_agent.py
+++ b/strix/agents/base_agent.py
@@ -134,7 +134,8 @@ class BaseAgent(metaclass=AgentMeta):
        }
        agents_graph_actions._agent_graph["nodes"][self.state.agent_id] = node

-        agents_graph_actions._agent_instances[self.state.agent_id] = self
+        with agents_graph_actions._agent_llm_stats_lock:
+            agents_graph_actions._agent_instances[self.state.agent_id] = self
        agents_graph_actions._agent_states[self.state.agent_id] = self.state

        if self.state.parent_id:
--- a/strix/telemetry/tracer.py
+++ b/strix/telemetry/tracer.py
@@ -799,17 +799,25 @@ class Tracer:
        )

    def get_total_llm_stats(self) -> dict[str, Any]:
-        from strix.tools.agents_graph.agents_graph_actions import _agent_instances
+        from strix.tools.agents_graph.agents_graph_actions import (
+            _agent_instances,
+            _completed_agent_llm_totals,
+            _agent_llm_stats_lock,
+        )
+
+        with _agent_llm_stats_lock:
+            completed_totals = dict(_completed_agent_llm_totals)
+            active_agents = list(_agent_instances.values())

        total_stats = {
-            "input_tokens": 0,
-            "output_tokens": 0,
-            "cached_tokens": 0,
-            "cost": 0.0,
-            "requests": 0,
+            "input_tokens": int(completed_totals.get("input_tokens", 0) or 0),
+            "output_tokens": int(completed_totals.get("output_tokens", 0) or 0),
+            "cached_tokens": int(completed_totals.get("cached_tokens", 0) or 0),
+            "cost": float(completed_totals.get("cost", 0.0) or 0.0),
+            "requests": int(completed_totals.get("requests", 0) or 0),
        }

-        for agent_instance in _agent_instances.values():
+        for agent_instance in active_agents:
            if hasattr(agent_instance, "llm") and hasattr(agent_instance.llm, "_total_stats"):
                agent_stats = agent_instance.llm._total_stats
                total_stats["input_tokens"] += agent_stats.input_tokens
--- a/strix/tools/agents_graph/agents_graph_actions.py
+++ b/strix/tools/agents_graph/agents_graph_actions.py
@@ -19,9 +19,55 @@ _running_agents: dict[str, threading.Thread] = {}

 _agent_instances: dict[str, Any] = {}

+_agent_llm_stats_lock = threading.Lock()
+
+
+def _empty_llm_stats_totals() -> dict[str, int | float]:
+    return {
+        "input_tokens": 0,
+        "output_tokens": 0,
+        "cached_tokens": 0,
+        "cost": 0.0,
+        "requests": 0,
+    }
+
+
+_completed_agent_llm_totals: dict[str, int | float] = _empty_llm_stats_totals()
+
 _agent_states: dict[str, Any] = {}


+def _snapshot_agent_llm_stats(agent: Any) -> dict[str, int | float] | None:
+    if not hasattr(agent, "llm") or not hasattr(agent.llm, "_total_stats"):
+        return None
+
+    stats = agent.llm._total_stats
+    return {
+        "input_tokens": stats.input_tokens,
+        "output_tokens": stats.output_tokens,
+        "cached_tokens": stats.cached_tokens,
+        "cost": stats.cost,
+        "requests": stats.requests,
+    }
+
+
+def _finalize_agent_llm_stats(agent_id: str, agent: Any) -> None:
+    stats = _snapshot_agent_llm_stats(agent)
+    with _agent_llm_stats_lock:
+        if stats is not None:
+            _completed_agent_llm_totals["input_tokens"] += int(stats["input_tokens"])
+            _completed_agent_llm_totals["output_tokens"] += int(stats["output_tokens"])
+            _completed_agent_llm_totals["cached_tokens"] += int(stats["cached_tokens"])
+            _completed_agent_llm_totals["cost"] += float(stats["cost"])
+            _completed_agent_llm_totals["requests"] += int(stats["requests"])
+
+            node = _agent_graph["nodes"].get(agent_id)
+            if node is not None:
+                node["llm_stats"] = stats
+
+        _agent_instances.pop(agent_id, None)
+
+
 def _is_whitebox_agent(agent_id: str) -> bool:
    agent = _agent_instances.get(agent_id)
    return bool(getattr(getattr(agent, "llm_config", None), "is_whitebox", False))
@@ -237,7 +283,7 @@ def _run_agent_in_thread(
        _agent_graph["nodes"][state.agent_id]["finished_at"] = datetime.now(UTC).isoformat()
        _agent_graph["nodes"][state.agent_id]["result"] = {"error": str(e)}
        _running_agents.pop(state.agent_id, None)
-        _agent_instances.pop(state.agent_id, None)
+        _finalize_agent_llm_stats(state.agent_id, agent)
        raise
    else:
        if state.stop_requested:
@@ -247,7 +293,7 @@ def _run_agent_in_thread(
        _agent_graph["nodes"][state.agent_id]["finished_at"] = datetime.now(UTC).isoformat()
        _agent_graph["nodes"][state.agent_id]["result"] = result
        _running_agents.pop(state.agent_id, None)
-        _agent_instances.pop(state.agent_id, None)
+        _finalize_agent_llm_stats(state.agent_id, agent)

        return {"result": result}

@@ -418,7 +464,8 @@ def create_agent(
        if inherit_context:
            inherited_messages = agent_state.get_conversation_history()

-        _agent_instances[state.agent_id] = agent
+        with _agent_llm_stats_lock:
+            _agent_instances[state.agent_id] = agent

        thread = threading.Thread(
            target=_run_agent_in_thread,
--- a/tests/telemetry/test_tracer.py
+++ b/tests/telemetry/test_tracer.py
@@ -10,6 +10,7 @@ from opentelemetry.sdk.trace.export import SimpleSpanProcessor, SpanExportResult
 from strix.telemetry import tracer as tracer_module
 from strix.telemetry import utils as telemetry_utils
 from strix.telemetry.tracer import Tracer, set_global_tracer
+from strix.tools.agents_graph import agents_graph_actions


 def _load_events(events_path: Path) -> list[dict[str, Any]]:
@@ -255,6 +256,75 @@ def test_events_with_agent_id_include_agent_name(monkeypatch, tmp_path) -> None:
    assert chat_event["actor"]["agent_name"] == "Root Agent"


+def test_get_total_llm_stats_includes_completed_subagents(monkeypatch, tmp_path) -> None:
+    monkeypatch.chdir(tmp_path)
+
+    class DummyStats:
+        def __init__(
+            self,
+            *,
+            input_tokens: int,
+            output_tokens: int,
+            cached_tokens: int,
+            cost: float,
+            requests: int,
+        ) -> None:
+            self.input_tokens = input_tokens
+            self.output_tokens = output_tokens
+            self.cached_tokens = cached_tokens
+            self.cost = cost
+            self.requests = requests
+
+    class DummyLLM:
+        def __init__(self, stats: DummyStats) -> None:
+            self._total_stats = stats
+
+    class DummyAgent:
+        def __init__(self, stats: DummyStats) -> None:
+            self.llm = DummyLLM(stats)
+
+    tracer = Tracer("cost-rollup")
+    set_global_tracer(tracer)
+
+    monkeypatch.setattr(
+        agents_graph_actions,
+        "_agent_instances",
+        {
+            "root-agent": DummyAgent(
+                DummyStats(
+                    input_tokens=1_000,
+                    output_tokens=250,
+                    cached_tokens=100,
+                    cost=0.12345,
+                    requests=2,
+                )
+            )
+        },
+    )
+    monkeypatch.setattr(
+        agents_graph_actions,
+        "_completed_agent_llm_totals",
+        {
+            "input_tokens": 2_000,
+            "output_tokens": 500,
+            "cached_tokens": 400,
+            "cost": 0.54321,
+            "requests": 3,
+        },
+    )
+
+    stats = tracer.get_total_llm_stats()
+
+    assert stats["total"] == {
+        "input_tokens": 3_000,
+        "output_tokens": 750,
+        "cached_tokens": 500,
+        "cost": 0.6667,
+        "requests": 5,
+    }
+    assert stats["total_tokens"] == 3_750
+
+
 def test_run_metadata_is_only_on_run_lifecycle_events(monkeypatch, tmp_path) -> None:
    monkeypatch.chdir(tmp_path)

--- a/tests/tools/test_agents_graph_whitebox.py
+++ b/tests/tools/test_agents_graph_whitebox.py
@@ -5,16 +5,24 @@ from strix.llm.config import LLMConfig
 from strix.tools.agents_graph import agents_graph_actions


-def test_create_agent_inherits_parent_whitebox_flag(monkeypatch) -> None:
-    monkeypatch.setenv("STRIX_LLM", "openai/gpt-5")
-
+def _reset_agent_graph_state() -> None:
    agents_graph_actions._agent_graph["nodes"].clear()
    agents_graph_actions._agent_graph["edges"].clear()
    agents_graph_actions._agent_messages.clear()
    agents_graph_actions._running_agents.clear()
    agents_graph_actions._agent_instances.clear()
+    agents_graph_actions._completed_agent_llm_totals.clear()
+    agents_graph_actions._completed_agent_llm_totals.update(
+        agents_graph_actions._empty_llm_stats_totals()
+    )
    agents_graph_actions._agent_states.clear()

+
+def test_create_agent_inherits_parent_whitebox_flag(monkeypatch) -> None:
+    monkeypatch.setenv("STRIX_LLM", "openai/gpt-5")
+
+    _reset_agent_graph_state()
+
    parent_id = "parent-agent"
    parent_llm = LLMConfig(timeout=123, scan_mode="standard", is_whitebox=True)
    agents_graph_actions._agent_instances[parent_id] = SimpleNamespace(
@@ -66,12 +74,7 @@ def test_create_agent_inherits_parent_whitebox_flag(monkeypatch) -> None:
 def test_delegation_prompt_includes_wiki_memory_instruction_in_whitebox(monkeypatch) -> None:
    monkeypatch.setenv("STRIX_LLM", "openai/gpt-5")

-    agents_graph_actions._agent_graph["nodes"].clear()
-    agents_graph_actions._agent_graph["edges"].clear()
-    agents_graph_actions._agent_messages.clear()
-    agents_graph_actions._running_agents.clear()
-    agents_graph_actions._agent_instances.clear()
-    agents_graph_actions._agent_states.clear()
+    _reset_agent_graph_state()

    parent_id = "parent-1"
    child_id = "child-1"
@@ -116,12 +119,7 @@ def test_delegation_prompt_includes_wiki_memory_instruction_in_whitebox(monkeypa
 def test_agent_finish_appends_wiki_update_for_whitebox(monkeypatch) -> None:
    monkeypatch.setenv("STRIX_LLM", "openai/gpt-5")

-    agents_graph_actions._agent_graph["nodes"].clear()
-    agents_graph_actions._agent_graph["edges"].clear()
-    agents_graph_actions._agent_messages.clear()
-    agents_graph_actions._running_agents.clear()
-    agents_graph_actions._agent_instances.clear()
-    agents_graph_actions._agent_states.clear()
+    _reset_agent_graph_state()

    parent_id = "parent-2"
    child_id = "child-2"
@@ -192,12 +190,7 @@ def test_agent_finish_appends_wiki_update_for_whitebox(monkeypatch) -> None:
 def test_run_agent_in_thread_injects_shared_wiki_context_in_whitebox(monkeypatch) -> None:
    monkeypatch.setenv("STRIX_LLM", "openai/gpt-5")

-    agents_graph_actions._agent_graph["nodes"].clear()
-    agents_graph_actions._agent_graph["edges"].clear()
-    agents_graph_actions._agent_messages.clear()
-    agents_graph_actions._running_agents.clear()
-    agents_graph_actions._agent_instances.clear()
-    agents_graph_actions._agent_states.clear()
+    _reset_agent_graph_state()

    parent_id = "parent-3"
    child_id = "child-3"
--- a/uv.lock
+++ b/uv.lock
@@ -158,7 +158,7 @@ wheels = [

 [[package]]
 name = "anthropic"
-version = "0.87.0"
+version = "0.86.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
    { name = "anyio" },
@@ -170,9 +170,9 @@ dependencies = [
    { name = "sniffio" },
    { name = "typing-extensions" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/d6/8f/3281edf7c35cbac169810e5388eb9b38678c7ea9867c2d331237bd5dff08/anthropic-0.87.0.tar.gz", hash = "sha256:098fef3753cdd3c0daa86f95efb9c8d03a798d45c5170329525bb4653f6702d0", size = 588982, upload-time = "2026-03-31T17:52:41.697Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/37/7a/8b390dc47945d3169875d342847431e5f7d5fa716b2e37494d57cfc1db10/anthropic-0.86.0.tar.gz", hash = "sha256:60023a7e879aa4fbb1fed99d487fe407b2ebf6569603e5047cfe304cebdaa0e5", size = 583820, upload-time = "2026-03-18T18:43:08.017Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/0d/02/99bf351933bdea0545a2b6e2d812ed878899e9a95f618351dfa3d0de0e69/anthropic-0.87.0-py3-none-any.whl", hash = "sha256:e2669b86d42c739d3df163f873c51719552e263a3d85179297180fb4fa00a236", size = 472126, upload-time = "2026-03-31T17:52:40.174Z" },
+    { url = "https://files.pythonhosted.org/packages/63/5f/67db29c6e5d16c8c9c4652d3efb934d89cb750cad201539141781d8eae14/anthropic-0.86.0-py3-none-any.whl", hash = "sha256:9d2bbd339446acce98858c5627d33056efe01f70435b22b63546fe7edae0cd57", size = 469400, upload-time = "2026-03-18T18:43:06.526Z" },
 ]

 [[package]]
@@ -2374,7 +2374,7 @@ name = "macholib"
 version = "1.16.4"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "altgraph", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "altgraph" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/10/2f/97589876ea967487978071c9042518d28b958d87b17dceb7cdc1d881f963/macholib-1.16.4.tar.gz", hash = "sha256:f408c93ab2e995cd2c46e34fe328b130404be143469e41bc366c807448979362", size = 59427, upload-time = "2025-11-22T08:28:38.373Z" }
 wheels = [
@@ -3815,7 +3815,7 @@ name = "pexpect"
 version = "4.9.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "ptyprocess", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "ptyprocess" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/42/92/cc564bf6381ff43ce1f4d06852fc19a2f11d180f23dc32d9588bee2f149d/pexpect-4.9.0.tar.gz", hash = "sha256:ee7d41123f3c9911050ea2c2dac107568dc43b2d3b0c7557a33212c398ead30f", size = 166450, upload-time = "2023-11-25T09:07:26.339Z" }
 wheels = [
@@ -4458,7 +4458,7 @@ name = "pyroscope-io"
 version = "0.8.16"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "cffi", marker = "sys_platform != 'win32'" },
+    { name = "cffi" },
 ]
 wheels = [
    { url = "https://files.pythonhosted.org/packages/a8/50/607b38b120ba8adad954119ba512c53590c793f0cf7f009ba6549e4e1d77/pyroscope_io-0.8.16-py2.py3-none-macosx_11_0_arm64.whl", hash = "sha256:e07edcfd59f5bdce42948b92c9b118c824edbd551730305f095a6b9af401a9e8", size = 3138869, upload-time = "2026-01-22T06:23:24.664Z" },
@@ -5439,7 +5439,7 @@ wheels = [

 [[package]]
 name = "strix-agent"
-version = "0.8.3"
+version = "0.8.2"
 source = { editable = "." }
 dependencies = [
    { name = "cvss" },
@@ -5511,7 +5511,7 @@ requires-dist = [
    { name = "rich" },
    { name = "scrubadub", specifier = ">=2.0.1" },
    { name = "tenacity", specifier = ">=9.0.0" },
-    { name = "textual", specifier = ">=6.0.0" },
+    { name = "textual", specifier = ">=4.0.0" },
    { name = "traceloop-sdk", specifier = ">=0.53.0" },
    { name = "uvicorn", marker = "extra == 'sandbox'" },
    { name = "xmltodict", specifier = ">=0.13.0" },
Author	SHA1	Message	Date
alex s	15c95718e6	fix: ensure LLM stats tracking is accurate by including completed subagents (#441 )	2026-04-13 00:09:13 -04:00
Ahmed Allam	62e9af36d2	Add Strix GitHub Actions integration tip	2026-04-12 12:43:41 -07:00