export — Data Export¶

Export bag data to AI/analytics-friendly formats.

Usage¶

# Parquet (default)
bagx export recording.db3

# JSON format
bagx export recording.db3 --format json

# AI-friendly mode (relative timestamps)
bagx export recording.db3 --ai --format parquet

# Specific topics only
bagx export recording.db3 --topics /gnss,/imu

# Custom output directory
bagx export recording.db3 --output ./my_export

Options¶

Flag	Description
`--format`, `-f`	`parquet` (default) or `json`
`--output`, `-o`	Output directory (default: `./export`)
`--topics`, `-t`	Comma-separated topic filter
`--ai`	Relative timestamps from bag start
`--flatten` / `--no-flatten`	Flatten nested fields (default: flatten)

AI mode¶

With --ai, timestamps are normalized to seconds from bag start (first message = 0.0). This is useful for ML pipelines where absolute epoch timestamps are not needed.

Output structure¶

One file per topic:

export/
  gnss.parquet          # /gnss topic
  imu.parquet           # /imu topic
  ouster_points.parquet # /ouster/points topic

Each file contains columns: timestamp_sec, timestamp_ns, plus flattened message fields.