Bench(refactor[lgtm]): Move the otel scripts into the stack they belong to

`scripts/lgtm/` holds the observability stack -- its compose files, dashboard
generator, telemetry wiring, and load driver. The smoke run and the acceptance
check sat outside it under an `otel_` prefix, though they are the two entry
points that stack exists to serve: `acceptance.py` already documented itself in
terms of `scripts.lgtm.generate_dashboards`, and `generate_dashboards` returns
the reference.

    scripts/otel_smoke.py      -> scripts/lgtm/smoke.py
    scripts/otel_acceptance.py -> scripts/lgtm/acceptance.py

Inside the directory the prefix was saying what the directory already says, so
it goes, matching `identity.py`, `telemetry.py`, and `load_tmux.py` beside
them.

Moving one level deeper broke two paths that had been counting directories,
both silently:

- `acceptance.py` derived the repository root with `parent.parent`, which now
  lands on `scripts/`, pointing every dashboard and helper lookup one level
  wrong.
- `smoke.py` put `__file__.parent / "lgtm"` on `sys.path` to import
  `telemetry`; from inside `lgtm/` that names a directory that does not exist.
  It now adds its own directory, the idiom `load_tmux.py` already uses.

`--help` used to say `otel_smoke.py`, which told a reader where to look. Bare
`smoke.py` does not, so both parsers name their path, as the orchestration
scripts do.

Verified: both run under the invocation their `just` recipes use, every
derived path resolves, and the lgtm tests pass.

Author: tony

CHANGES

@@ -286,7 +286,7 @@ engine instrumentation seam, so the exporters are ordinary sinks and libtmux
 itself gains no OpenTelemetry dependency.
 
 The Grafana dashboards under `scripts/lgtm/dashboards/` are generated rather
-than hand-edited, and `scripts/otel_acceptance.py` runs each panel's own query
+than hand-edited, and `scripts/lgtm/acceptance.py` runs each panel's own query
 and fails naming any panel that came back empty. See `scripts/lgtm/README.md`.
 
 Every run is stamped with its branch, revision, repository, worktree, and an

justfile

@@ -173,17 +173,17 @@ otel-dashboards:
 # Drive a real tmux workload through the engine seam into LGTM
 [group: 'otel']
 otel-smoke *args:
-    uv run --group otel python scripts/otel_smoke.py {{ args }}
+    uv run --group otel python scripts/lgtm/smoke.py {{ args }}
 
 # Verify every dashboard panel's own queries return data
 [group: 'otel']
 otel-acceptance *args:
-    uv run --group otel python scripts/otel_acceptance.py {{ args }}
+    uv run --group otel python scripts/lgtm/acceptance.py {{ args }}
 
 # Start the stack, run the workload, then verify every panel end to end
 [group: 'otel']
 otel-verify:
-    uv run --group otel python scripts/otel_acceptance.py --start-stack --smoke
+    uv run --group otel python scripts/lgtm/acceptance.py --start-stack --smoke
 
 # Drive the engines under a load shape (ramping arrival rate) via rampa
 [group: 'otel']

scripts/lgtm/README.md

@@ -82,7 +82,7 @@ $ just otel-ports
 
 ## What the workload emits
 
-`scripts/otel_smoke.py` runs every transport — subprocess and control mode,
+`scripts/lgtm/smoke.py` runs every transport — subprocess and control mode,
 sync and async — against a throwaway tmux server, and emits four signals:
 
 Metrics are `tmux_requests_total`, `tmux_commands_total`, `tmux_inlined_total`,
@@ -273,7 +273,7 @@ mid-run -- `destroy-unattached off` only survives *detach*.
 A dashboard that renders is not a dashboard that works. A panel whose query
 returns nothing looks exactly like a panel reporting a healthy zero.
 
-`scripts/otel_acceptance.py` reads the generated JSON, expands the template
+`scripts/lgtm/acceptance.py` reads the generated JSON, expands the template
 variables the way Grafana would, runs every panel's own query against
 Prometheus, Loki, Tempo, or Pyroscope, and fails naming any panel that returned
 nothing. Because it reads the dashboards themselves, a panel added to the

scripts/otel_acceptance.py scripts/lgtm/acceptance.pyscripts/otel_acceptance.py renamed to scripts/lgtm/acceptance.py

@@ -28,7 +28,7 @@
 import urllib.parse
 import urllib.request
 
-ROOT = pathlib.Path(__file__).resolve().parent.parent
+ROOT = pathlib.Path(__file__).resolve().parents[2]
 DASHBOARDS = ROOT / "scripts" / "lgtm" / "dashboards"
 SERVICE = "libtmux-engines"
 
@@ -327,7 +327,9 @@ def check_until(endpoints: Endpoints, timeout: float, poll: float) -> list[Resul
 
 def main(argv: list[str] | None = None) -> int:
     """Run the acceptance sweep and print a per-panel report."""
-    parser = argparse.ArgumentParser(description=__doc__)
+    parser = argparse.ArgumentParser(
+        prog="scripts/lgtm/acceptance.py", description=__doc__
+    )
     parser.add_argument("--prometheus", default="http://127.0.0.1:9099")
     parser.add_argument("--loki", default="http://127.0.0.1:3100")
     parser.add_argument("--tempo", default="http://127.0.0.1:3200")
@@ -336,7 +338,7 @@ def main(argv: list[str] | None = None) -> int:
         "--start-stack", action="store_true", help="run scripts/lgtm/up.sh first"
     )
     parser.add_argument(
-        "--smoke", action="store_true", help="run scripts/otel_smoke.py first"
+        "--smoke", action="store_true", help="run scripts/lgtm/smoke.py first"
     )
     parser.add_argument(
         "--settle",
@@ -362,7 +364,7 @@ def main(argv: list[str] | None = None) -> int:
         subprocess.run([str(ROOT / "scripts" / "lgtm" / "up.sh")], check=True)
     if args.smoke:
         subprocess.run(
-            [sys.executable, str(ROOT / "scripts" / "otel_smoke.py")], check=True
+            [sys.executable, str(ROOT / "scripts" / "lgtm" / "smoke.py")], check=True
         )
     if args.start_stack or args.smoke:
         time.sleep(args.settle)

scripts/lgtm/generate_dashboards.py

@@ -9,8 +9,8 @@
 start and ``tests/test_lgtm_dashboards.py`` fails if the committed copy differs,
 so the two cannot silently diverge.
 
-Every panel must be backed by telemetry ``scripts/otel_smoke.py`` actually
-emits. ``scripts/otel_acceptance.py`` executes each panel's own queries and
+Every panel must be backed by telemetry ``scripts/lgtm/smoke.py`` actually
+emits. ``scripts/lgtm/acceptance.py`` executes each panel's own queries and
 fails on any that returns nothing, which is what keeps a board honest.
 """
 

scripts/lgtm/load_tmux.py

@@ -1,6 +1,6 @@
 """Load-shape the tmux engines with rampa, and export the result to LGTM.
 
-``otel_smoke.py`` answers "does telemetry flow" by running flat out for a fixed
+``smoke.py`` answers "does telemetry flow" by running flat out for a fixed
 duration. That is the wrong shape for asking where a transport stops keeping
 up, because a closed loop of N workers slows down with the system: offered load
 falls as latency rises, and the graph bends politely instead of breaking.

scripts/otel_smoke.py scripts/lgtm/smoke.pyscripts/otel_smoke.py renamed to scripts/lgtm/smoke.py

@@ -28,7 +28,7 @@
 import typing as t
 import uuid
 
-sys.path.insert(0, str(pathlib.Path(__file__).parent / "lgtm"))
+sys.path.insert(0, str(pathlib.Path(__file__).resolve().parent))
 
 import telemetry
 
@@ -46,7 +46,7 @@
 from libtmux.experimental.engines.instrumentation import CountingSink
 from libtmux.server import Server
 
-logger = logging.getLogger("libtmux.otel_smoke")
+logger = logging.getLogger("libtmux.lgtm.smoke")
 
 PLAIN = CommandRequest.from_args("list-panes", "-a", "-F", "#{pane_id}")
 LISTING = CommandRequest.from_args("list-windows", "-a", "-F", "#{window_id}")
@@ -253,7 +253,7 @@ def lane_totals(counts: CountingSink) -> dict[str, int]:
 
 def main(argv: list[str] | None = None) -> int:
     """Run every lane under full telemetry and print the local counts."""
-    parser = argparse.ArgumentParser(description=__doc__)
+    parser = argparse.ArgumentParser(prog="scripts/lgtm/smoke.py", description=__doc__)
     parser.add_argument("--run-id", default=f"smoke-{uuid.uuid4().hex[:8]}")
     parser.add_argument(
         "--spike",

tests/test_lgtm_dashboards.py

@@ -1,7 +1,7 @@
 """Structural contracts for the generated Grafana dashboards.
 
 These run offline. They cannot tell whether a panel has data -- that is
-``scripts/otel_acceptance.py`` against a live stack -- but they do keep the
+``scripts/lgtm/acceptance.py`` against a live stack -- but they do keep the
 committed JSON honest about its generator and about the datasources it binds
 to, which is where a board rots silently.
 """
@@ -135,7 +135,7 @@ def test_acceptance_expands_every_template_variable() -> None:
     through as text and match nothing.
     """
     spec = importlib.util.spec_from_file_location(
-        "otel_acceptance", _ROOT / "scripts" / "otel_acceptance.py"
+        "lgtm_acceptance", _ROOT / "scripts" / "lgtm" / "acceptance.py"
     )
     assert spec is not None
     assert spec.loader is not None