CueLight vs Vocily AI

CueLight

CueLight gives a voice pipeline a local end-of-turn probability every 20 ms from audio alone, with a 1.9 MB model and Apache-2.0 weights. It is one day old, English only, installs from Git rather than PyPI, has no CI or changelog, and its accuracy figures are the author's own.

Vocily AI

The REST API is documented in detail for a product a few months old, with an OpenAPI 3.1 file, required idempotency keys on calls, a Retry-After header on 429 and signed webhooks. No status page, SLA, SDK or security programme was found, and two releases in September 2026 removed fields under /v1 on the day they were announced.

CueLight

1. Feed 320 float32 samples at 16 kHz mono per push. Resample first, or the call raises an AssertionError
2. Spell the preset as snappy or careful. Any other name silently uses default thresholds
3. Call clear() before a new call or session. Pass reset_on_turn_start=False if your pipeline already resets per turn
4. Pin weights by downloading them to a local folder and passing that path to from_pretrained
5. Don't use it to identify speakers or clone voices. The training data's conditions forbid it

Vocily AI

1. Send a fresh UUID in Idempotency-Key on every POST /v1/calls. The API refuses the request without it
2. Pass a from_number the workspace holds. The shared Test Number returns 403 from the API
3. Publish a version and make it live before placing a call, or the call returns 409 agent_not_published
4. Branch on detail.code, falling back to the top-level code. A 429 body is flat, so read Retry-After
5. Read currency on every call and chat record. Costs are in rupees or dollars depending on the organisation