JiuwenSwarm is an intelligent AI Agent built on openJiuwen. It extends the powerful capabilities of large language models directly to your fingertips through various communication apps you use daily.
Python
"""Fresh-evidence gate for unified telemetry topology and the 21 metrics."""
from __future__ import annotations
import os
from pathlib import Path
from typing import Any
import pytest
from scripts.e2e_telemetry_trace import (
DEFAULT_EVIDENCE,
EXPECTED_METRICS,
FORBIDDEN_SPAN_NAMES,
FORBIDDEN_SPAN_PREFIXES,
EvidenceError,
load_fresh_evidence,
)
pytestmark = [pytest.mark.integration, pytest.mark.system]
@pytest.fixture(scope="module")
def evidence() -> dict[str, Any]:
path = Path(os.getenv("TELEMETRY_E2E_EVIDENCE", str(DEFAULT_EVIDENCE)))
max_age = float(os.getenv("TELEMETRY_E2E_MAX_AGE_SECONDS", "900"))
try:
return load_fresh_evidence(path, max_age_seconds=max_age)
except EvidenceError as error:
pytest.fail(str(error))
def _flow_spans(evidence: dict[str, Any], flow: dict[str, Any]) -> list[dict[str, Any]]:
request_id = flow["request_id"]
return [
span
for span in evidence["spans"]
if (span.get("attributes") or {}).get("jiuwenswarm.request.id") == request_id
]
def test_streaming_and_non_streaming_websocket_flows_are_fresh(
evidence: dict[str, Any],
) -> None:
success = {
(flow["mode"], flow["streaming"])
for flow in evidence["flows"]
if flow.get("scenario") == "success"
and flow.get("transport") == "websocket"
and flow.get("terminal_event") != "chat.error"
}
assert ("code.normal", True) in success
assert ("agent.plan", False) in success
started = evidence["run_started_unix_nano"]
assert all(flow["started_unix_nano"] >= started for flow in evidence["flows"])
def test_gateway_and_agent_roots_share_trace_and_parent_chain(
evidence: dict[str, Any],
) -> None:
all_spans = evidence["spans"]
by_trace_and_id = {(span["trace_id"], span["span_id"]): span for span in all_spans}
assert len(by_trace_and_id) == len(all_spans), "duplicate exported span identity"
for flow in evidence["flows"]:
if (
flow.get("scenario") != "success"
or flow.get("entrypoint") != "gateway"
):
continue
spans = _flow_spans(evidence, flow)
gateways = [span for span in spans if span["name"] == "channel.request"]
assert len(gateways) == 1
roots = [
span
for span in spans
if span["name"].startswith("agent.")
and span.get("parent_span_id") == gateways[0]["span_id"]
and (span.get("attributes") or {}).get("jiuwenswarm.mode")
in (
{"agent", flow["mode"]}
if flow["mode"] == "agent.plan"
else {flow["mode"]}
)
]
assert len(roots) == 1
assert gateways[0]["trace_id"] == roots[0]["trace_id"]
assert roots[0]["parent_span_id"] == gateways[0]["span_id"]
for span in all_spans:
parent_id = span.get("parent_span_id")
if parent_id:
assert (span["trace_id"], parent_id) in by_trace_and_id, (
f"orphan span {span['name']} parent={parent_id}"
)
def test_core_spans_are_enriched_without_enterprise_duplicates(
evidence: dict[str, Any],
) -> None:
names = [span["name"] for span in evidence["spans"]]
assert "llm.call" in names
assert any(name.startswith("tool.") for name in names)
assert not FORBIDDEN_SPAN_NAMES.intersection(names)
assert not any(name.startswith(FORBIDDEN_SPAN_PREFIXES) for name in names)
llm = next(
span
for span in evidence["spans"]
if span["name"] == "llm.call"
and "gen_ai.usage.input_tokens" in span["attributes"]
)
attrs = llm["attributes"]
assert attrs["gen_ai.request.model"]
assert attrs["gen_ai.usage.input_tokens"] >= 0
assert attrs["gen_ai.span.type"] == "model"
assert attrs["jiuwenswarm.request.id"]
def test_all_21_metrics_have_points_from_this_run(evidence: dict[str, Any]) -> None:
started = evidence["run_started_unix_nano"]
fresh_names = {
point["name"]
for point in evidence["metrics"]
if point["time_unix_nano"] >= started
}
assert EXPECTED_METRICS <= fresh_names
assert len(EXPECTED_METRICS) == 21