wuji-cli-logs

Mental Model

• Logs are local text files written by Wuji SDK, Studio, and stderr capture, stored under ~/.wuji/logs/. The CLI discovers and processes these files — it does not create them.
• Three text sources: sdk_*.log, studio_*.log, stderr_*.log. Device communication dumps (device_*.bin) are excluded by default and gated behind --with-dump.
• Export produces a single ZIP containing the selected log files, a host snapshot (host_snapshot.json), doctor diagnosis results (doctor_diagnosis.json — environment checks under env, device discovery and per-device checks under device), device snapshots + flash history (devices/<sn>_<ts>_device.json / _flash.jsonl, when devices are found), matching hand model calibration recordings, and a manifest (manifest.json). The snapshot and diagnosis are included even when no log files are found.
• Calibration recordings are automatic: runs from ~/.wuji/calibration/recordings/ that intersect the same inclusive local-date range are copied under calibration-recordings/<run_id>/. Partial or legacy runs with no ended_at are filtered as a single point at their local started_at date.
• Recording manifest compatibility: new hand-model runs use schema_version: 2 and calibration_type: "hand_model"; export also accepts legacy schema 1 / "ik" runs and preserves every source manifest unchanged.
• Redaction is layered: credentials (JWT, Bearer Token, API Key, Token, License, Password, AWS Key) are always redacted; private data (username, user paths, hostname, SSID, IPv4, MAC) is redacted by default and can be disabled via --no-redact (internal builds only). Calibration .mcap files contain raw motion data and are never redacted.
• Exit codes: 0 = success; 1 = any failure, including device diagnosis, log export, calibration recording, or device snapshot/flash collection failures. Recoverable items remain listed in the bundle manifest.

Subcommands