Batch Processor

Overview

The Batch Processor is an API that performs post-processing jobs on your call media.

It accepts a recording ID from a Daily meeting or a URL to a video/audio file (e.g. mp4, mp3) and can produce a transcript of the audio, summary of the transcript, or a SOAP note generated from the audio.

For a complete list and description of each endpoint, refer to the Batch Processor API reference.

Types of jobs

There are few predefined post-processing jobs that the batch processor can perform from the audio of a media file:

  1. Generate a transcript
  2. Generate a SOAP note
  3. Generate a summary
  4. Generate a summary, with your own custom prompt (contact us for access to custom prompts)

Post-call transcripts

The batch processor can generate transcripts from the audio of a media file. The transcription is powered by Deepgram.

Example transcripts

Transcripts are formatted as txt, srt, vtt, and json.

Text:

Speaker 0: Good afternoon. My name is Anass Ali. I'm one of the final year medical students. Before we start, can I check your name and age
Speaker 1: Yeah? It's, Zaines sand I'm fifty five years old.
Speaker 0: And how would you like me to address you today?
Speaker 1: Zaines is fine. It's
Speaker 0: a pleasure to meet you, Zaines. I've just been asked by one of the doctors just to have a conversation with you really, but was brought you into hospital. Anything we discussed remain confidential within the medical team as well. Are you happy to go ahead?

SRT:

1
00:00:09,975 --> 00:00:34,350
Good afternoon. My name is Anass Ali. I'm one of the final year medical students. Before we start, can I check your name and age Yeah? It's, Zaines sand I'm fifty five years old. And how would you like me to address you today? Zaines is fine. It's a pleasure to meet you, Zaines. I've just been asked by one of the doctors just to have a conversation with you really, but was brought you into hospital. Anything we discussed remain confidential within the medical team as well. Are you happy to go ahead? Yeah. Absolutely. Okay. Great. So tell me a bit about what's been going on.
2
00:00:35,129 --> 00:00:38,270
Just this morning as I was leaving my house, but in the morning,
3
00:00:38,875 --> 00:00:41,054
I experienced some pretty bad chest pain.

VTT:

WEBVTT
NOTE
Transcription provided by Deepgram
Request Id: 5534e5b9-4779-43d0-9428-1bca5976ce79
Created: 2023-09-12T07:18:26.162Z
Duration: 579.5469
Channels: 1
1
00:00:09.975 --> 00:00:34.350
- Good afternoon. My name is Anass Ali. I'm one of the final year medical students. Before we start, can I check your name and age Yeah? It's, Zaines sand I'm fifty five years old. And how would you like me to address you today? Zaines is fine. It's a pleasure to meet you, Zaines. I've just been asked by one of the doctors just to have a conversation with you really, but was brought you into hospital. Anything we discussed remain confidential within the medical team as well. Are you happy to go ahead? Yeah. Absolutely. Okay. Great. So tell me a bit about what's been going on.
2
00:00:35.129 --> 00:00:38.270
- Just this morning as I was leaving my house, but in the morning,

JSON:

Medical SOAP notes

A SOAP note is a particular type of documentation that summarizes an encounter between a patient and medical staff.

The batch processor begins by generating a transcript from the audio of a media file. It feeds the transcript (along with optional metadata concepts from ScienceIO) into Daily’s private HIPAA-compliant GPT-4 AI model to generate a SOAP note.

The use of metadata concepts from ScienceIO is optional and can be enabled or disabled. SOAP notes can be generated in either case.

Example SOAP note

The SOAP note is written as a json file. An example file looks like this (this is fake data):

Post-call summaries

For use-cases where the SOAP format isn't required, the batch processor can generate summaries from the audio of a media file.

Example summary

The summary is written as a text file. An example file looks like this (this is fake data):

Medical student Anass Ali interviews a 55-year-old patient named Zaines who was admitted to the hospital after experiencing severe chest pain. Zaines describes the pain as sudden, lasting for 40 minutes, and felt like pressure on his chest. The pain also radiated down his arm and up his neck. He rates the pain as an 8 or 9 out of 10. Zaines is worried that he might be having a heart attack, as his father had a similar experience when he was 65. He also reveals that he has high blood pressure and high cholesterol, for which he takes Ramipril and a statin, respectively. Zaines has been a smoker for 25 years and drinks alcohol on the weekends. The medical team plans to discuss Zaines' symptoms and conduct further examinations and investigations.

Feature configuration

These configuration options are optional domain properties. They affect every batch processing job on a Daily domain.

By default, batch processor outputs are stored with Daily's HIPAA-compliant storage and SOAP note generation uses medical concepts.

Property NameTypeDescriptionDefault
batch_processor_bucketobjectDefines a custom S3 bucket where the batch processor will write its output to.null (output stored at Daily)
batch_processor_enable_medical_conceptbooleanEnables whether the batch processor will use ScienceIO metadata concepts when generating SOAP notes.true

This is the same style configuration as setting up your own custom S3 storage for recordings. Please note: if recordings_bucket property was set on either the domain or room configuration when a recording was made, then when passing in the resulting recording id to a batch processor POST request, the domain configuration for recordings_bucket must match the recording configuration.

API reference

For a complete list and description of each endpoint, refer to the Batch Processor API reference.