Blowhorn Download

Configuration reference

Precedence, every .blowhorn.yaml option, environment variables, and output logging.

Configuration reference

How Blowhorn resolves what a run uses, and where its output goes. blowhorn/config_schema.py is the one list of options; blowhorn config show prints each with its stored, environment, effective, and source values. blowhorn config set edits the file without rewriting it: comments, order, and quoting survive.

Precedence, highest first: the CLI flag, the environment variable, the .blowhorn.yaml default, then the built-in. The file lives beside the current directory by default (./.blowhorn.yaml); --config points at another one.

Every option .blowhorn.yaml supports

Default key Environment Built-in Means
defaults.profile BLOWHORN_PROFILE all The profile a run uses when --profile is not given
defaults.platform BLOWHORN_PLATFORM all The platform a run targets when --platform is not given
defaults.exclude BLOWHORN_EXCLUDE none Profiles left out of --profile all
defaults.headless BLOWHORN_HEADLESS off Unattended background window
defaults.headed BLOWHORN_HEADED off Visible window in front
defaults.dry_run BLOWHORN_DRY_RUN off Simulate every run
defaults.pace BLOWHORN_PACE normal fast, normal, slow, or a number; commands that never pause accept and ignore it
defaults.no_mouse_move BLOWHORN_NO_MOUSE_MOVE off Skip the pointer simulation before clicks
defaults.logs.directory BLOWHORN_LOG_DIR logs/ in the checkout Where daily logs go
defaults.logs.retention_days BLOWHORN_LOG_RETENTION_DAYS 30 Days of daily logs kept; 0 keeps all
defaults.chrome.launch_mode BLOWHORN_CHROME_LAUNCH extension How a run reaches Chrome
defaults.chrome.automation_data_dir BLOWHORN_CHROME_AUTOMATION_DATA_DIR Blowhorn's own directory Where a launched Chrome keeps its session
defaults.chrome.data_dir BLOWHORN_CHROME_DATA_DIR Your real Chrome directory Where the installed Chrome is read from
defaults.slack.auth_mode BLOWHORN_SLACK_AUTH your captured session first Which Slack credential may send: session, user-token, bot-token, or a list tried in order
(flag only) BLOWHORN_TEE_STDOUT off Flush the stdout copy per write for a supervising wrapper
defaults.store.backend BLOWHORN_STORE_BACKEND postgres The one store backend; anything else is refused by name
defaults.store.organization BLOWHORN_STORE_ORGANIZATION none; while unset, nothing runs The organization every row is scoped to
defaults.store.config BLOWHORN_STORE_CONFIG a private file outside the checkout The private document holding the connection

headed and headless are one choice with two spellings: the higher layer wins, so a --headless flag beats a headed default from the file, and --headed wins a tie. The settings screen reports each on its own; a flagless run resolves them together so exactly one survives.

A version-1 file that still says persistent reads as extension on every run; only a validated config set of another key moves the file to version 2. The ten store.* connection options are private: they stay out of config show.

Output logging

The daily log is a copy; the terminal streams are never dropped, piped or not. The directory resolves as --log-dir, then BLOWHORN_LOG_DIR, then defaults.logs.directory, then the checkout's logs/. A default that cannot be used fails silently; a directory you named reports why. Only files the retention pass recognizes by name are ever deleted: daily logs past their days (30 by default), and the three dateless capture files rolling past their size bound. Per-job logs, the ledger, and the scheduler heartbeat live beneath the same directory and are never pruned by it.

--tee-stdout is about liveness, not logging: it flushes the stdout copy per write, and with no usable log directory it line-buffers the real stdout instead. It is read before command parsing, so it is not a per-command flag.

Last modified October 9, 2026: docs: sync from master (5f4e718)