vibium_testersai
The easiest way to AI-upgrade your existing Vibium suite. You already have the tests — this adds AI Checks at the moments that matter, without replacing your framework, runner, or CI pipeline.
analyze_* call, or define your own.TestersAISession(session) wraps a Vibium session and duck-types against its current console/network API.
# Download the bundle from the Downloads page, then:
pip install vibium
pip install ./testersai-python-vibium-0.1.0.tar.gz
A complete, runnable Python + Vibium example.
from vibium import Session
from vibium_testersai import TestersAISession
with Session() as s:
ta = TestersAISession(s)
ta.capture_console()
ta.capture_network()
s.goto("https://example.com")
ta.analyze_screenshot()
ta.analyze_console()
ta.analyze_network()
Runs under any Python test framework — pytest, unittest, plain script. Default disk sink writes JSON. If the AI call fails — rate limit, hang, firewall, no network — the SDK gives up fast and returns a skipped result. Your Vibium test is never blocked.
Rule of thumb: wherever a human reviewer would pause to look during a manual run. These recommendations are tuned for Vibium — pick the ones that fit your suite.
The page just re-rendered. Ask "does this look right?" before any interaction.
Login, add-to-cart, toggle, submit. The UI just reflected a new state — where regressions hide.
You were about to check one thing. Ask the AI about everything else for free.
API returned, spinner gone, toast shown. Catch broken empty-states and stale data.
A single console + network check at the end of every test catches issues your assertions ignored.
Mobile vs. desktop, light vs. dark, locale change. One call per viewport.
You can also point the open SDK at a screenshot produced by vibium go on the command line, or by Vibium's MCP server — same analyze_* surface.
Or grab the monolith ZIP (all languages, all adapters).