Deepgram audio intelligence brief
Synchronous Deepgram audio intelligence workflow. It transcribes one explicit media URL, extracts utterances and channel alternatives when returned, and summarizes transcript quality, summary, topics, intents, sentiment, speakers, and follow-up actions. It never uploads files, opens streaming sockets, configures voice agents, generates tokens, uses callbacks, or calls arbitrary Deepgram routes.
How it works
Inspect the data fetches, transforms, gates, and output this recipe runs.
Process (10 steps)
normalized_url
trim
Trim whitespace from setting audio url
transcription
?
utterances_raw
default
Apply default
channels
default
Apply default
utterances
slice
Take a subset of
channel_count
count
Count items in channels
utterance_count
count
Count items in utterances raw
summarize
?
utterance_table
to_table
Format results as a data table
card
to_summary
Format results as a summary card
Settings
Configurable at install. Defaults shown — change them anytime in Recued.
focus
setting
=
Summarize the transcript, Deepgram summary, sentiment, topics, intents, speaker turns, quality signals, and follow-up actions. Use only the Deepgram response.
model
setting
=
nova-3
topics
setting
=
true
intents
setting
=
false
deepgram
setting
=
language
setting
=
en
audio url
setting
=
row limit
setting
=
20
sentiment
setting
=
true
summarize
setting
=
v2
max length
setting
=
900
Trust & control
What installing this recipe would let it do. Recued grants these permissions at install — review them there before approving.