Datasets:

ppak10
/

Agentic-SLS-Telemetry

"""Shared paths and helpers for the extract scripts.
Raw data lives in this dataset's own source/ tree (source-based layout,
2026-07-16); extract scripts read source/ and write to data/.
"""
import json
from pathlib import Path

ROOT = Path(__file__).parent.parent
DATA_DIR = ROOT / "data"

# Since 2026-07-16 this dataset is source-based: raw inputs live in OUR OWN
# source/ tree (written by the recorder repo's `sls-export` /
# `sls-deliver-frames`). data/exports in the recorder repo is retired.
# AGENTIC_ROOT still points at the containing recorder repo (submodule,
# two climbs) — used only for the transient frame buffer of builds not yet
# delivered into source/frames/ zips.
AGENTIC_ROOT = ROOT.parent.parent
SOURCE_DIR = ROOT / "source"
EXPORTS_DIR = SOURCE_DIR / "recorder"   # builds/events/frames jsonl + sensors.csv
TELEMETRY_DIR = SOURCE_DIR / "telemetry"      # {NNN}.parquet raw ticks
POSITION_DIR = SOURCE_DIR / "position_hf"     # {NNN}.parquet
FRAMES_BUFFER = AGENTIC_ROOT / "data" / "frames"  # pre-delivery loose frames


def iter_jsonl(path: Path):
    """Stream a JSONL file row-by-row. Skips blank lines."""
    with path.open(encoding="utf-8") as f:
        for line in f:
            line = line.strip()
            if not line:
                continue
            yield json.loads(line)


def load_build_to_profile_name() -> dict[int, str]:
    """Read upstream events.jsonl, return {build_id: printProfileName}.
    Uses the `build_start` event's `payload.printProfileName` field. Builds
    that never logged a build_start event (the recorder joined mid-run) are
    omitted from the map.
    """
    events_path = EXPORTS_DIR / "events.jsonl"
    out: dict[int, str] = {}
    for ev in iter_jsonl(events_path):
        if ev.get("kind") != "build_start":
            continue
        name = (ev.get("payload") or {}).get("printProfileName")
        if name:
            out[ev["build_id"]] = name
    return out

	"""Shared paths and helpers for the extract scripts.

	Raw data lives in this dataset's own source/ tree (source-based layout,
	2026-07-16); extract scripts read source/ and write to data/.
	"""
	import json
	from pathlib import Path

	ROOT = Path(__file__).parent.parent
	DATA_DIR = ROOT / "data"

	# Since 2026-07-16 this dataset is source-based: raw inputs live in OUR OWN
	# source/ tree (written by the recorder repo's `sls-export` /
	# `sls-deliver-frames`). data/exports in the recorder repo is retired.
	# AGENTIC_ROOT still points at the containing recorder repo (submodule,
	# two climbs) — used only for the transient frame buffer of builds not yet
	# delivered into source/frames/ zips.
	AGENTIC_ROOT = ROOT.parent.parent
	SOURCE_DIR = ROOT / "source"
	EXPORTS_DIR = SOURCE_DIR / "recorder" # builds/events/frames jsonl + sensors.csv
	TELEMETRY_DIR = SOURCE_DIR / "telemetry" # {NNN}.parquet raw ticks
	POSITION_DIR = SOURCE_DIR / "position_hf" # {NNN}.parquet
	FRAMES_BUFFER = AGENTIC_ROOT / "data" / "frames" # pre-delivery loose frames


	def iter_jsonl(path: Path):
	"""Stream a JSONL file row-by-row. Skips blank lines."""
	with path.open(encoding="utf-8") as f:
	for line in f:
	line = line.strip()
	if not line:
	continue
	yield json.loads(line)


	def load_build_to_profile_name() -> dict[int, str]:
	"""Read upstream events.jsonl, return {build_id: printProfileName}.

	Uses the `build_start` event's `payload.printProfileName` field. Builds
	that never logged a build_start event (the recorder joined mid-run) are
	omitted from the map.
	"""
	events_path = EXPORTS_DIR / "events.jsonl"
	out: dict[int, str] = {}
	for ev in iter_jsonl(events_path):
	if ev.get("kind") != "build_start":
	continue
	name = (ev.get("payload") or {}).get("printProfileName")
	if name:
	out[ev["build_id"]] = name
	return out