Migrating to v1.0.0

v1.0.0 changes the recording format and tidies the Python and C++ surfaces. This guide shows how to move existing recordings, scripts, and code; the complete change list is in the v1.0.0 release notes.

You are affected if you have recordings made before v1.0.0, scripts that call the shipped tools, or code on the v0.x Python or C++ APIs. Update the whole toolchain together — pre-v1.0.0 tools open v1.0.0 recordings as empty, with no error explaining why.

Recordings

Convert each pre-v1.0.0 recording once:

Visualizer: open the old recording — the visualizer offers the conversion, then opens the converted file.

Command line (scriptable) — see voyant_recording_migrate:

voyant_recording_migrate --input my_capture.bin
# writes my_capture.vynt next to the input
# --output <path> to choose the destination, --force to replace an existing file

New recordings use the .vynt extension, and the recorder enforces it — output paths ending in anything else are rejected. Recording to a path that already exists now fails unless you opt in (--force on voyant_logger_binary, overwrite=True in Python, the overwrite config field in C++).

Old raw-peak CSV captures have no converter — re-capture, or reach out to Voyant support about updating existing captures.

Command-line tools

The shipped tools dropped their carbon qualifier (voyant_carbon_client_check → voyant_client_check, voyant_carbon_simulator → voyant_simulator, and so on). Update scripts to the new names; the Retired Tools page maps every old name.

The CSV converters are now internal-only and no longer documented (see Retired Tools). For analysis exports, use the Python bindings’ points() matrix — the Python table below maps the old CSV column names to their points() equivalents — or the voyant-ros repository.

Python

v0.x	v1.0.0
`points()` / `valid_points()` — `(N, 7)` float32: `x, y, z, radial_vel, snr_linear, nanosecs_since_frame, drop_reason`	`(N, 12)` float64, spherical by default: `range_m, azimuth_rad, elevation_rad, doppler_mps, snr, calibrated_reflectance, timestamp_nanosecs, azimuth_idx, elevation_idx, drop_reason, combine_method, user_data`. `points(cartesian=True)` swaps the first three columns to `x, y, z`.
`points_extended()`, `special_test()`, `spherical()` accessor families	Removed — select columns from `points()` by name instead
`xyzv()` fourth column `radial_vel`	Renamed `doppler_mps` (same data)
`valid_mask()` returns `list[bool]`	Returns a NumPy boolean array
`describe()` prints	Returns a string
`VoyantPlayback(filter_points=...)`	`keep_invalid_points=...`
PCD helpers’ `valid_only=...`	`keep_invalid_points=...`
`config.set_keep_invalid_points(...)`	`config.set_diagnostic_mode(...)`
`VoyantRecorder(buffer_size_mb=...)`	Removed (the buffer sizes itself); `overwrite=` added
`start_peak_dump()` / `stop_peak_dump()` / `is_peak_dumping()`	Diagnostic capture — see below
`VoyantClient`	Removed — use `CarbonClient`
`voyant_api.utils` compatibility shim	Removed
`set_missing_elevations()`, `set_interp_range_threshold()`, `set_interp_doppler_threshold()`	Removed (elevation interpolation is gone; it was off by default)

Column-name-keyed access is the order-proof way to consume and edit points:

df = pd.DataFrame(frame.points(), columns=frame.points_columns())
edited = frame.with_points(df[frame.points_columns()].to_numpy())

drop_reason codes are renumbered in v1.0.0 — treat 1 as valid and any other value as dropped, and do not carry v0.x code meanings forward.

Peak dump → diagnostic capture

The peak-dump API is replaced by diagnostic capture, which writes the raw peak stream to a .vynt sidecar while the point cloud keeps running (the old dump required the client to be stopped):

config.set_diagnostic_mode(True)  # before client.start()
client.start_diagnostic_capture("capture_peaks.vynt")
# record frames alongside with VoyantRecorder — together they form the support bundle
client.stop_diagnostic_capture()  # requests the stop; returns immediately
while client.is_diagnostic_capturing():  # wait before reading the file
    time.sleep(0.1)

C++

v0.x	v1.0.0
`bool tryReceiveFrame()` + `VoyantFrameWrapper& latestFrame()`	`std::optional<VoyantFrame> tryReceiveFrame()` — the frame is the new native `VoyantFrame` class: `points()`, `xyz()`, `sensorState()`, `hostState()`
`VoyantClient`, `PointsClient`	Removed — use `CarbonClient`
`VoyantPlayback(..., bool filter_points = false)`	`VoyantPlayback(..., bool keep_invalid_points = false)` — invalid points are now removed by default; pass `true` to keep them
`CarbonConfig::setKeepInvalidPoints(...)`	`setDiagnosticMode(...)`
`startPeakDump()` / `stopPeakDump()` / `isPeakDumping()`	`startDiagnosticCapture(path)` / `stopDiagnosticCapture()` / `isDiagnosticCapturing()`, with `setDiagnosticMode(true)` set before `start()`. `stopDiagnosticCapture()` returns immediately — poll `isDiagnosticCapturing()` before reading the file
Recorder config `bufferSizeMb`	Removed (the buffer sizes itself); `overwrite` field added
`setMissingElevations()`, `setInterpRangeThreshold()`, `setInterpDopplerThreshold()`	Removed (elevation interpolation is gone; it was off by default)
C FFI `voyant_playback_is_file_open`	Removed, no C++ replacement (Python keeps `is_open`)
C FFI `voyant_recorder_create`	Gained error output parameters

← Release notes
v1.0.0