AI-generated Answer for sari_33713
Short answer: not directly. ElevenLabs Agents don’t have a built‑in way for a tool call to automatically include the “audio so far” as a parameter.
Supported ways to get call audio to your system:
- Post‑call webhook: Enable post_call_audio webhooks to receive the full conversation audio as base64 after the call ends. This is the simplest way to deliver audio to downstream services. (elevenlabs.io)
- Fetch by conversation_id after the call: Have the tool receive the system__conversation_id as an input (via Dynamic Variables), then your server can call GET /v1/convai/conversations/{conversation_id}/audio to download the recording. You can check has_audio/has_user_audio/has_response_audio with GET /v1/convai/conversations/{conversation_id}. (elevenlabs.io)
Important constraints:
- Audio must be saved: If “Audio saving” is disabled for the agent, recordings aren’t stored, so the audio download endpoint won’t have data; transcripts can still exist. Ensure audio saving is enabled if you plan to fetch audio. (elevenlabs.io)
- Zero Retention Mode (ZRM): When ZRM is enabled, call recordings aren’t stored. To retrieve call information in ZRM scenarios, use post‑call webhooks. (elevenlabs.io)
If you need mid‑call audio: the docs do not describe an API to fetch a partial/ongoing recording during the call; the documented delivery mechanisms are post‑call webhooks and the conversation audio download for completed recordings. (elevenlabs.io)
Summary: You can’t attach “audio so far” to a tool call natively. Use post_call_audio webhooks, or pass conversation_id to your tool and fetch the recording afterward (with audio saving on, and not in ZRM). (elevenlabs.io)
Sources:
Post-call webhooks | ElevenLabs Documentation
Dynamic variables | ElevenLabs Documentation
Audio saving | ElevenLabs Documentation
Zero Retention Mode (per-agent) | ElevenLabs Documentation
Post-call webhooks | ElevenLabs Documentation
Post-call webhooks | ElevenLabs Documentation
vagrant prism
swift gull