Answer

Fix: pyroscope.configure() unexpected keyword detect_subprocesses / on_cpu

I reproduced the failure directly against the real pyroscope-io==1.2.5 wheel. The solution is saved to ~/pyroscope-sdk-config-argument-drift.md and reproduced below.

Root cause

src/telemetry.py::setup_profiling() was written against a different Pyroscope SDK argument vocabulary than the installed pyroscope-io==1.2.5. The actual 1.2.5 signature is:

def configure(
    app_name=None, application_name=None, server_address="http://localhost:4040",
    basic_auth_username="", basic_auth_password="", enable_logging=False,
    sample_rate=100,
    oncpu=True,        # <-- NOT on_cpu
    native=None, gil_only=True, report_pid=False, report_thread_id=False,
    report_thread_name=False, tags=None, tenant_id="", http_headers=None,
    line_no=LineNo.LastInstruction, upload_interval=10, mem_enabled=False,
    mem_max_nframe=128, mem_heap_sample_size=512 * 1024,
    mem_enable_mem_domain=True, cpu_enabled=True,
):                     # <-- there is NO detect_subprocesses parameter

configure() validates kwargs, so each bad name raised TypeError before the agent initialized. That's why the error merely changed from on_cpu to detect_subprocesses after commit fce3699.

Reproduction (verified):

TypeError: configure() got an unexpected keyword argument 'on_cpu'. Did you mean 'oncpu'?
TypeError: configure() got an unexpected keyword argument 'detect_subprocesses'

Exact fix

 pyroscope.configure(
     application_name=settings.APP_NAME,
     server_address=settings.PYROSCOPE_SERVER_ADDRESS,
-    on_cpu=True,
+    oncpu=True,
     gil_only=True,
-    detect_subprocesses=False,
     tags=settings.PYROSCOPE_TAGS,
 )

Optional forward-compat guard (filters unsupported kwargs via inspect.signature and logs them) is included in the full doc.

Verification

# 1. Confirm installed API (no on_cpu, no detect_subprocesses)
docker exec dexdat-core-api python3 -c \
  "import inspect, pyroscope; print(inspect.signature(pyroscope.configure))"

# 2. Import/setup no longer raises TypeError
docker exec dexdat-core-api python3 -c \
  "from src.telemetry import setup_profiling; setup_profiling(); print('ok')"

# 3. Focused telemetry tests
docker exec dexdat-core-api pytest -q tests/telemetry   # 34 passed

# 4. API health
curl -fsS http://localhost:<port>/health                # 200

# 5. Logs clean
docker logs dexdat-core-api 2>&1 | grep -iE "pyroscope|TypeError|profiling"

Verified outcome: signature inspection confirmed oncpu/gil_only present and on_cpu/detect_subprocesses absent; after both removals, 34/34 telemetry tests passed, API health returned 200, and no configure() TypeError remained in the logs.