Stream commaCarSegments CAN data¶

Goal¶

List comma.ai commaCarSegments route segments for a vehicle platform, plan a bounded replay, and stream CAN frames from a selected openpilot rlog.zst.

Prerequisites¶

Network access to the HuggingFace commaai/commaCarSegments dataset.
Optional openpilot LogReader/cereal support installed for actual comma-rlog frame streaming (uv pip install git+https://github.com/commaai/openpilot.git; requires Python 3.12.x).
A bounded plan (--limit, --max-frames, or --max-seconds) before working with large segment lists.

List Tesla segments¶

Start with a platform filter and a small manifest limit:

canarchy datasets replay catalog:comma-car-segments \
  --platform TESLA_MODEL_3 \
  --list-files \
  --limit 20 \
  --json

The result contains stable file id values. Use one with --file in later commands.

Dry-run a selected segment¶

Dry-run resolves metadata without opening the rlog payload stream:

canarchy datasets replay catalog:comma-car-segments \
  --platform TESLA_MODEL_3 \
  --file 0 \
  --dry-run \
  --json

Stream frames¶

Stream a bounded JSONL sample once optional parser support is available:

canarchy datasets replay catalog:comma-car-segments \
  --platform TESLA_MODEL_3 \
  --file 0 \
  --format jsonl \
  --max-frames 1000

Pipe directly into analysis commands when using candump output:

canarchy datasets replay catalog:comma-car-segments \
  --platform TESLA_MODEL_3 \
  --file 0 \
  --max-frames 1000 \
  | canarchy stats --file - --json

Stream a local rlog¶

If you already downloaded an rlog.zst, stream it as a local dataset file:

canarchy datasets stream ./rlog.zst \
  --source-format comma-rlog \
  --format jsonl \
  --provider-ref catalog:comma-car-segments \
  --max-frames 1000

If optional openpilot parser support is missing, CANarchy returns COMMA_RLOG_SUPPORT_UNAVAILABLE; list and dry-run workflows still work.