Home/Docs/Retell

Capture from Retell

Multichannel recording export is built in: capture pulls the call’s per-party recording by call id and scores it offline, preferring the PII-scrubbed file.

The recording you capture feeds a five-dimension conversation-test: separated caller and agent channels are what the Conversation (turn-taking) and Speech (timing) dimensions read, and the same file drops into hotato test run beside a transcript and trace for the Outcome and Policy checks.

Stack	Automatic pull	Separate channels	Contract-grade default	Network use
Vapi	Yes: `capture` fetches by call id	Yes, out of the box	Yes	One call: fetch the recording
Twilio	Yes: `capture` fetches by recording SID	No by default: must request `RecordingChannels=dual` at creation	No, opt-in	One call: fetch the recording
Retell	Yes: `capture` fetches by call id	Yes, once call recording is turned on	Opt-in: enable recording first	One call: fetch the recording
LiveKit	No: egress runs in your own infra	Yes, if egress records both tracks	Opt-in: configure egress for two tracks	None from Hotato
Pipecat	No: the `setup` recorder runs in your pipeline	Yes, the scaffold writes caller ch0 / agent ch1	Yes, once the recorder is added	None: local recorder, no network call

Automatic pull: does hotato capture fetch it for you. Separate channels: is caller/agent separation what you get. Contract-grade default: is that separation on without extra setup. Network use: what capture calls out to, not your infrastructure.

Enable call recording

Turn on call recording for your agent; Retell exposes per-party (multichannel) recordings on the call object after the call ends.
Grab the call id from the dashboard or the call webhook payload.

Score it: auto-pull

Set RETELL_API_KEY + call id: capture fetches the recording (only network call: that download), scores offline. --stack retell = fix knobs; --expect yield/hold = interruption vs. backchannel.

bash

export RETELL_API_KEY=...
uvx hotato capture --stack retell --call-id <call-id>

Under the hood: GET /v2/get-call/<call-id>, then scrubbed_recording_multi_channel_url (preferred) or recording_multi_channel_url, each party on its own channel.

The plain recording_url is a mono mix; Hotato rejects it unless you pass --allow-mono: indicative only; no contract-grade talk-over attribution.

Same call, two exports. The multi-channel file keeps caller and agent on separate channels, so an overlap has an owner. recording_url flattens both onto one mono track: the same overlap can’t be attributed. --allow-mono opts out of that gate.

Or score a WAV you already have

Downloaded the multichannel file? Point Hotato at it: caller on channel 0, agent on channel 1.

bash

uvx hotato run --stereo call.wav --stack retell --expect yield

Channels flipped? Pass --caller-channel 1 --agent-channel 0.

Pulling more than one call?

hotato connect retell --api-key <key> stores the key once, locally (0600); Retell has no list endpoint to check it against up front, so the check runs on the first pull instead. From there, Hotato pulls exactly the calls you name by --call-id: pull bulk-fetches those recordings, and sweep pulls, then ranks the worst turn-taking moments across them in one offline dashboard. Every stack’s pull support →

bash

uvx hotato pull --stack retell --call-id c1 --call-id c2
uvx hotato sweep --stack retell --call-id c1 --call-id c2

The dual-channel gate

Separated caller and agent audio is what makes overlap attributable; speech activity comes from exposed VAD thresholds, frame-inspectable. The scorer takes one two-channel WAV or two aligned mono WAVs.