latimer-woods-tech/Word Is Bond — Voice Agent Test

Test a voice agent you run against a scenario suite in CI and fail the build on a score or outcome regression.

Trust Signals

name	description	required	default
api-key	Word Is Bond API key (wb_live_…). Store it as a repository secret and pass it here. Get one free at https://studio.wordis-bond.com.	yes	—
suite-id	The id of the scenario suite to run (created in the Studio or via POST /api/tests/suites).	yes	—
target	The agent under test — a system you run. A SIP URI (sip:agent@host), a WebSocket URL (wss://…), a hosted agent id (e.g. wb-demo-dental), or a (paid, pro+) E.164 number (+14155550123). Runs SIP-direct by default so CI minutes stay on the near-free tier.	yes	—
threshold	Minimum passing score (0–100). If set, the run fails when the score is below it. Enforced regardless of fail-on.	no	—
baseline	A prior run id to compare against for regression gating. Defaults to the previous run in the suite trend history.	no	—
fail-on	What to gate on: "regression" (default, score drop vs baseline), "threshold" (score below threshold), or "outcome" (Outcome Verification assertions).	no	regression
goal	Optional goal override passed to the synthetic caller.	no	—
language	Optional language override (default: en).	no	—
api-url	Override the API origin (for staging). Defaults to https://api.wordis-bond.com.	no	—
timeout	Maximum seconds to wait for a live run to complete before failing. Default 600.	no	600
poll-interval	Seconds between run-status polls. Default 5.	no	5

name	description
score	The aggregate (worst-case) score of the run, 0–100.
passed	"true" if the gate passed, "false" otherwise.
delta	Score change vs the baseline (regression gate only), or empty when no baseline exists.
run-id	The id of the primary run.
report-url	The shareable, tokenized public report URL for the primary run.