# Logging Use the process logger for diagnosable failures and meaningful state transitions. Skip noise (render churn, routine success paths, prop dumps). ## Which logger | Process | Import | API | |---------|--------|-----| | Renderer / remote | `import { logger } from '/@/renderer/utils/logger'` | `logger.debug\|info\|warn\|error(message, meta?)` | | Main | `import log from '/@/main/logger'` | `log.debug\|info\|warn\|error(...)` | Do not use bare `console.*` for new logging in app code — the loggers format, debounce (renderer), forward to electron-log / files, and respect the configured level. ## When to log Add a log when the code path is hard to reconstruct from the UI alone: - Failures and catch blocks that would otherwise vanish (`error` / `warn`) - Engine / connection / auth / playback transitions that change behavior (`info` or `debug`) - Fallback paths (e.g. playback engine fell back to web) (`warn` / `info`) Skip: hot loops, every React render, expected empty states, data the user already sees as a toast unless the log adds process-side detail. ## Levels Configured threshold is `debug` | `info` (`LogLevel` in `/@/shared/logger/types`). Message severity is `debug` | `info` | `warn` | `error`. | Severity | Use for | |----------|---------| | `error` | Failed operation; include the error in `meta` or args | | `warn` | Degraded / unexpected but continuing | | `info` | Notable lifecycle (engine chosen, reconnect, update check) | | `debug` | Detail useful when digging (payload shapes, step traces) — gated off unless level is `debug` | ## Message shape - First arg: short human message string. - Optional second arg (renderer): structured `meta` object — prefer `{ error, … }` over string-concatenating errors. - No secrets: passwords, tokens, full auth headers, or anything `sanitize-for-diagnostics` would strip from exports. Main diagnostic packaging uses `/@/shared/utils/sanitize-for-diagnostics` — keep ad-hoc logs equally clean.