Readiness diagnostics
vibium ready checks your local browser installation and tests configured AI.
It reports setup problems and fixes before you start browser work.
vibium readyExample with an installed browser and no AI settings:
Installed: chrome (stable) — /path/to/chrome[PASSED] browser.installation: chrome (stable) and its required executable files are installed.[SKIPPED] browser.connection: Browser launch and BiDi connectivity are not tested by readiness.[SKIPPED] ai: AI is not configured; optional for direct browser automation. Fix: For Run, Check, or find by description, run vibium setup ai, then vibium ready ai.
READY: Browser installation checks passed; launch and BiDi connectivity were not tested.No AI settings file yet. Run vibium setup ai, or vibium config init and edit ~/.config/vibium/ai.env.No AI settings means plain ready skips AI. Partially configured or failing AI
produces a failure. Run and Check require AI even when browser-only readiness
passes. Readiness does not assess your application’s behavior.
Choose what to test
Section titled “Choose what to test”| Command | Requested checks |
|---|---|
vibium ready | Selected browser installation and AI if configured |
vibium ready browser | Selected/default browser installation; AI is ignored |
vibium ready browser chrome | Chrome executable files |
vibium ready browser firefox --channel beta | Firefox beta executable files |
vibium ready ai | AI configuration and provider tool round-trip; no browser required |
vibium ready ai anthropic --model your-model | Anthropic using that model and ANTHROPIC_API_KEY |
vibium ready speech | The say and listen providers, with one phrase sent through both; no browser required |
Run readiness during initial setup, after changing configuration, or when troubleshooting. You do not need to run it before every browser command.
Browser readiness
Section titled “Browser readiness”The browser check uses Vibium’s existing path resolution to list installations across Chrome and Firefox channels. It checks that the selected browser and required driver files exist and are executable. It never launches a browser or driver. Other listed installations are informational.
This keeps setup checks lightweight and avoids browser startup triggering OS
or GPU initialization. Browser launch and BiDi connectivity are reported as
skipped, rather than claimed to work based on file checks.
Selection uses --engine / VIBIUM_ENGINE, --channel /
VIBIUM_ENGINE_CHANNEL, and the existing version/path settings. A positional
browser overrides the environment’s engine. When that changes the engine,
the channel resets to the new engine’s default unless --channel is explicit.
Conflicting positional and --engine selections are errors.
Chrome uses Vibium’s cached Chrome for Testing and matching driver;
VIBIUM_ENGINE_PATH supports a custom Firefox executable.
Readiness does not install browsers, start a daemon, use your active session, or inspect a remote browser. A missing installation produces an explicit fix:
vibium install --engine firefox --channel betavibium ready browser firefox --channel betaThe second command checks the installed files. Passing does not prove that the browser can start, connect over BiDi, or run your application. Use the same engine/channel for subsequent live work.
AI readiness
Section titled “AI readiness”AI checks validate shared VIBIUM_AI_* settings, authentication, model access,
tool calling, and a structured response using the same provider adapters as
Run and Check. They make at most two model requests with a 60-second budget;
API charges may apply. Invalid settings skip the provider request.
Use per-call model options
with ready or ready ai: --provider, --model, --ai-base-url, and
--reasoning-effort. ready ai <provider> is equivalent to selecting that
provider with --provider. Changing provider requires an explicit model and
clears inherited endpoint/effort defaults. A conflicting provider argument
and flag is an error. No per-call option changes later defaults.
Vibium loads ~/.config/vibium/ai.env for empty AI variables. Nonempty process
environment still wins. The report ends by naming the file it loaded, or by
pointing at vibium setup ai when there is none, so you can see where a
setting comes from and how to change it. Credentials and raw provider error
bodies are not displayed.
ready ai works with no browser installed and ignores browser settings.
Use it before checking an archive.
It does not test screenshot support or application behavior.
Speech readiness
Section titled “Speech readiness”vibium ready speechready speech tests vibium say and vibium listen.
It picks the providers as those commands do, has the say provider render
one phrase into a WAV file, and has the listen provider transcribe that
file.
[PASSED] say: piper-server en_US-bryce-medium rendered 3.7 s of speech in 5.5 s.[PASSED] listen: sherpa-server heard "the quick brown fox jumps over the lazy dog" in 2.8 s.
READY: Say and listen passed: a phrase went through both providers.sayfails when no provider is configured, its engine or key is missing, or the render fails. The message carries the error code thatvibium saywould give, such asnot_configuredorengine_failed.listenfails the same way, and also when what it heard is not the phrase. It is skipped whensayrendered nothing for it to hear.- Both checks must pass. Plain
vibium readydoes not test speech.
Nothing is played or captured, so speakers, the microphone, a browser page
and a device are not tested. A hosted provider is asked for one render and
up to two transcripts, which may cost money. The render does not go into
the cache vibium say uses.
JSON and exit status
Section titled “JSON and exit status”vibium ready --jsonvibium ready ai --jsonvibium ready speech --jsonWithout --json, output is readable text. With it, stdout contains one JSON
object using the normal {ok, result, error?} CLI envelope:
result.ready: whether all requested checks passedresult.scope:all,browser,ai, orspeechresult.summary: concise readiness explanationresult.checks:{name, status, message, fix?}entriesresult.notes: limitations and setup guidanceresult.browsers: discovered browser selections when browser checks ran; entries includeengine,channel,installed,selected, and resolvedpath/driverwhen available
Individual check statuses are passed, failed, or skipped. A missing
optional AI configuration is skipped only by plain ready.
Exit 0 means the requested checks passed; exit 1 means setup needs
attention. Thus ready ai fails when AI is unconfigured, while plain ready
can succeed when browser installation checks pass. Browser launch and
connectivity remain untested. Neither result is a Check verdict.