Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.sf-voice.sh/llms.txt

Use this file to discover all available pages before exploring further.

Telnyx is a communications platform that provides programmable voice, SIP trunking, and call control through its REST API and WebSocket-based Media Streaming. If your voice AI infrastructure runs on Telnyx, Mise connects via media streaming and call control webhooks to index acoustic features across your full call corpus.
Mise is in private alpha. Your Mise media streaming endpoint, webhook URL, and API key are provided after your team is granted access. Request access.

​
What Mise captures from Telnyx

For each Telnyx call, Mise captures:
  • Per-turn audio: Both inbound and outbound audio, segmented by speaker and indexed for acoustic features
  • Call control events: Call initiated, answered, bridged, transferred, and hung-up events from Call Control webhooks
  • Call metadata: Call UUID, connection ID, caller ID, direction, and timestamps
  • Acoustic features per turn: Tone, prosody, frustration signal, interruptions, and silence patterns
If your pipeline generates transcripts via Telnyx AI features or a third-party STT provider, Mise can ingest those alongside the audio to enrich search results.

​
How to connect Telnyx to Mise

1

Request alpha access

Request access to Mise. After onboarding, you will receive your Mise WebSocket endpoint for media streaming and your webhook URL for call control events.
2

Enable media streaming on your call

Use the Telnyx Call Control API to start media streaming on an active call. Point the streaming WebSocket to your Mise ingest endpoint.
import requests

TELNYX_API_KEY = "your-telnyx-api-key"
CALL_CONTROL_ID = "your-call-control-id"

# Mise endpoint — provided after alpha access
MISE_WS_ENDPOINT = "wss://ingest.mise.sh/v1/stream/telnyx"
MISE_API_KEY = "your-mise-api-key"

response = requests.post(
    f"https://api.telnyx.com/v2/calls/{CALL_CONTROL_ID}/actions/streaming_start",
    headers={
        "Authorization": f"Bearer {TELNYX_API_KEY}",
        "Content-Type": "application/json",
    },
    json={
        "stream_url": f"{MISE_WS_ENDPOINT}?api_key={MISE_API_KEY}",
        "stream_track": "both_tracks",
        "enable_dialogflow": False,
    },
)
3

Configure call control webhooks

In your Telnyx Mission Control Portal, set the webhook URL for your connection or phone number to your Mise call events endpoint. This allows Mise to receive call lifecycle events and correlate them with your audio stream.Set your Telnyx Webhook URL to:
https://ingest.mise.sh/v1/webhook/telnyx?api_key=YOUR_MISE_API_KEY
Mise listens for the following Telnyx event types:
  • call.initiated
  • call.answered
  • call.hangup
  • streaming.started
  • streaming.stopped
4

Verify ingest

Place a test call. After the call ends, open the Mise dashboard and confirm the call appears in your corpus. Run a test query to verify per-turn acoustic features are indexed.

​
Streaming both tracks

Setting stream_track to both_tracks is required for Mise to separate and index each speaker’s turns independently. If you set stream_track to inbound_track or outbound_track only, Mise will index that single track but cannot detect interruptions or overlap events across both speakers.
If you start media streaming in response to a call.answered webhook event, make sure to start the stream before significant audio has been exchanged. Mise indexes from the point streaming begins — it cannot retroactively index audio that was not forwarded.
Telnyx media streams use mulaw encoding at 8000 Hz by default. Your Mise endpoint is configured to accept this format. Do not re-encode the audio before forwarding it.