Greenroom.

A practice-run recording goes in. A coaching bundle comes out: deterministic filler counts with timestamps, pacing analysis per segment, weak-phrase rewrites, cue cards for the live show, and a 0-100 energy/clarity score. The coaching bundle is itself anchored to the Knox event chain plus OpenTimestamps Bitcoin proof — bundle integrity verifies from the bundle bytes alone.

A 50-second voice-narrated tour.

Narration synthesized by macOS Premium neural voice (Ava); screen capture rendered headlessly against this same live page via Playwright. No human voice, no human image — fully synthetic, fully reproducible.

Four measurements + cue cards, anchored.

The coaching pipeline is deterministic where the signal is measurable, and Sonnet-driven where the signal is qualitative — and every artifact in both classes is hash-anchored before it reaches the bundle.

One primitive, many surfaces.

The Anchored Receipt Engine is the substrate: a recording in, a transcript + structured extraction + Knox anchor + OpenTimestamps proof + FRE 902(13)/(14)-styled PDF certification template (operator-fill) out. Greenroom is one skin that consumes a receipt bundle from the engine and produces a coaching bundle. The engine ships once; skins compose atop it.

Two receipts, two anchors, two verifiable claims: the recording was processed exactly as represented, and the coaching analysis was produced exactly as represented. Both verify from their bundle bytes alone.

Try the engine.

The Greenroom coaching pipeline is exposed as a single bearer-gated endpoint at the Anchored Receipt Engine. Hand it a receipt bundle from POST /v1/receipts and it returns the coaching bundle.

# 1. Emit a receipt bundle from a recording
curl -X POST https://arc.bonissystems.com/v1/receipts \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -F "recording=@meeting.mp4"
# → returns: { bundle: { ... } }

# 2. Pass that bundle to the Greenroom coaching pipeline
curl -X POST https://arc.bonissystems.com/v1/coaching \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"bundle": <paste-bundle-from-step-1>}'
# → returns: { coachingBundle: { ... } }

The coaching bundle JSON returned by /v1/coaching verifies independently — every artifact hash, every Knox anchor, and every embedded .ots proof can be re-derived from the bundle bytes alone. Hand it to any reviewer and they can check it without contacting the engine again.

The verifier never contacts Bonis.

Each coaching bundle includes the full integrity surface: canonical-JSON manifest, per-artifact SHA-256 hashes, Knox anchor receipts (sequence number, payload hash, verification URL), and embedded OpenTimestamps proof bytes. The standalone verifier CLI ships with the engine package and runs offline.

With --bitcoin-walk the verifier additionally decodes embedded .ots proofs, walks each to a Bitcoin block header against any Bitcoin node, and returns 0 only when all anchors confirm. Bonis Systems is not in the verification path.

A real receipt + coaching bundle, packaged for download.

The archive below contains the three artifacts from a 41-second rehearsal recording — generated via macOS say for the demo: the audio file itself, the receipt bundle returned by POST https://arc.bonissystems.com/v1/receipts, and the coaching bundle returned by POST https://arc.bonissystems.com/v1/coaching. A README inside explains each file. Drop either JSON bundle into the in-browser verifier just below to recompute every per- artifact hash and the bundle manifest hash locally; no Bonis runtime contact required.

Drop a bundle. Hashes are recomputed locally.

The verifier below ports the engine's canonical-JSON algorithm and runs it against the file you drop. SHA-256 is computed via the browser's Web Crypto API. The bundle never leaves the page; nothing is uploaded to Bonis. For each check the declared hash is compared to the value recomputed locally — bit-perfect or fail. Try it on the sample bundle above, or on any bundle the engine has previously emitted.

The page that explains anchored receipts can produce one of itself.

Click the button below and the browser builds a marker text — visit timestamp, user-agent string, 16-byte nonce — hashes those bytes via Web Crypto, and submits only the SHA-256 digest to the Knox public-anchor endpoint. The response is a Knox sequence number, an anchor hash, and a self-authenticating affidavit that you can download. The marker text never leaves your browser; only the digest does.

Greenroom is infrastructure, not a webinar platform.

Greenroom does not replace LiveKit, Zoom, Riverside, or Loom. It does not record the live show, host the audience, or run the broadcast. It does not retain the uploaded recording on Bonis infrastructure beyond the duration of the receipt-emission request — every request is stateless and the engine never sees user identity beyond the bearer token used to authenticate. Embedders are expected to front the engine with their own identity layer (SSO / IAP / VPC ingress). The Greenroom skin produces a coaching bundle from a recording; everything else is upstream of the primitive.

Common questions.