Best answer

Integrating Fireflies.ai and ChatGPT


I am trying to capture the full transcript from Fireflies.ai and send it into a chat gpt prompt. Zapier is not extracting the full transcript from Fireflies.ai. It only allows you to leverage the transcript url or file from Fireflies.ai. How can ensure that the full transcript is captured within this workflow? 

Chat GPT Prompt:

Summarize this meeting transcript and include main topics, action items, next steps and detailed notesFire Flies.ai inserted data elements:{{180316144__note[]caption__speaker_name}}{{180316144__note[]caption__sentence}}{{180316144__note[]note}}{{180316144__summary}}
icon

Best answer by SamB 26 January 2023, 11:22

View original

This post has been closed for comments. Please create a new post if you need help or have a question about this topic.

11 replies

Userlevel 7
Badge +11

Thanks so much for sharing that helpful alternative here, @courtneywaid. It’s much appreciated! 😁

Userlevel 1

Hi @E_glenn! Not sure how tied you are to Fireflies.ai, but I recently came across this tool that returns the text from a transcript that you can then feed into OpenAI: https://whisperapi.com/

 

You will need to send your file or file URL via a webhook step, and it returns the text directly in the response that you can use in your following action steps. 

 

Hope that helps! I was recently exploring the same challenge with OpenAI and Fireflies and came across your post. 

Userlevel 7
Badge +11

Hey @E_glenn! 👋

I did some checking and it definitely looks like we don’t get the content of the transcripts, only a url to the transcript file in the transcriptUrl field. I can’t see any existing feature requests to get the transcript file contents. I’d recommend contacting our Support team to get a feature request opened up on your behalf. You can reach them here: https://zapier.com/app/get-help

In the meantime, Troy’s suggested workarounds of uploading the file to an app like Dropbox to extract its content or connect to Fireflies.ai’s API via a Webhooks by Zapier action might do the trick. If you have success with those workarounds or find a different solution here please do let us know!

Userlevel 7
Badge +14

@E_glenn 

You can try using the Fireflies.ai API: https://docs.fireflies.ai/

App APIs can be used in Zaps with the Webhooks app as an action step: https://zapier.com/apps/webhook/help

(See screenshots below)

 

OR

 

You may have to upload the Transcript to Dropbox to get the file contents:

 

###

 

1. {{180316144__title}}

2. {{180316144__audioUrl}}

3. {{180316144__date}}

4. {{180316144__prettyDate}}

5. {{180316144__transcriptFileUrl}}

6. {{180316144__attendees}}

7. {{180316144__transcriptUrl}}

8. {{180316144__meetingQuestions}}

9. {{180316144__meetingTaskNotes}}

10. {{180316144__meetingMetricNotes}}

11. {{180316144__mainAttendee}}

12. {{180316144__meetingDateTimeNotes}}

13. {{180316144__organizerUrl}}

14. {{180316144__host}}

15. {{180316144__id}}

16. {{180316144__meetingPricingNotes}}

17. {{180316144__attendeeEmailNames}}

18. {{180316144__note[]caption__speaker_name}}

19. {{180316144__note[]action}}

20. {{180316144__note[]caption__algScore}}

21. {{180316144__note[]caption__averageConfidence}}

22. {{180316144__note[]caption__endTime}}

23. {{180316144__note[]caption__filterType}}

24.{{180316144__note[]caption__id}}

25. {{180316144__note[]caption__importance}}

26. {{180316144__note[]caption__importanceScore}}

27. {{180316144__note[]caption__incoming}}

28. {{180316144__note[]caption__index}}

29. {{180316144__note[]caption__match}}

30. {{180316144__note[]caption__metrics}}

31. {{180316144__note[]caption__metrics[]category}}

32. {{180316144__note[]caption__metrics[]word}}

33. {{180316144__note[]caption__questionFilter}}

34. {{180316144__note[]caption__readability}}

35. {{180316144__note[]caption__sentence}}

36. {{180316144__note[]caption__sentiment}}

37. {{180316144__note[]caption__sentimentType}}

38. {{180316144__note[]caption__speaker}}

39. {{180316144__note[]caption__speaker_id}}

40. {{180316144__note[]caption__task}}

41. {{180316144__note[]caption__taskMatch}}

42. {{180316144__note[]caption__time}}

43. {{180316144__note[]id}}

44. {{180316144__note[]note}}

45. {{180316144__note[]question}}

46. {{180316144__mainAttendeeEmailName__displayName}}

47. {{180316144__mainAttendeeEmailName__email}}

48. {{180316144__summary}}

these are all the available data elements from the trigger stage 

Userlevel 7
Badge +14

@E_glenn 

Was seeking screenshots that shows all the data points returned from Step 1, like these.

 

What do you mean when you say “Full list of data points”? The below details consist of the test trigger from Fireflies.ai. I can see all the data points that are sent from Fireflies.ai into Zapier within the test:

 

id

uFku9T9fGu

date

2023-01-23T19:35:00.000Z

prettyDate

Mon Jan 23 at 7:35 PM UTC

title

Notion test 2

host

eric.glenn@phytonconsulting.com

mainAttendee

eric.glenn@phytonconsulting.com

mainAttendeeEmailName

email

eric.glenn@phytonconsulting.com

displayName

eric glenn

attendees

null

attendeeEmailNames

transcriptUrl

https://app.fireflies.ai/view/Notion-test-2::uFku9T9fGu

audioUrl

https://cdn.fireflies.ai/uFku9T9fGu/audio.mp3

transcriptFileUrl

https://download-ff.s3.amazonaws.com/uFku9T9fGu-Timestamp-Speakers-Notiontest2st.docx

note

1

id

0

note

I have a prompt that features some test data, but it's not the correct.

action

false

question

false

caption

sentence

I have a prompt that features some test data, but it's not the correct.

speaker_id

2

time

26.1

endTime

33.04

speaker

2

averageConfidence

0.9591864285714288

speaker_name

Eric Glenn

index

4

match

none

filterType

none

incoming

false

metrics

1

word

prompt

category

Nouns

2

word

data

category

Nouns

3

word

correct

category

Nouns

sentiment

0

sentimentType

neutral

questionFilter

false

importance

0.5701988935470581

readability

0.8505799770355225

task

0.9648336172103882

algScore

20.097169126455018

importanceScore

10.455891370773315

taskMatch

false

id

4

2

id

1

note

I'm wondering why it is set up like that.

action

false

question

false

caption

sentence

I'm wondering why it is set up like that.

speaker_id

1

time

36.62

endTime

40.01

speaker

1

averageConfidence

0.8229922222222223

speaker_name

Eric Glenn

index

6

match

none

filterType

none

incoming

false

metrics

sentiment

0.2222222222222222

sentimentType

positive

questionFilter

false

importance

0.11034409701824188

readability

0.6480569243431091

task

0.933576226234436

algScore

18.069248179320383

importanceScore

9.443268376588822

taskMatch

false

id

6

3

id

2

note

We have an automation test with Barrick and Fred.

action

true

question

false

caption

sentence

We have an automation test with Barrick and Fred.

speaker_id

1

time

51.03

endTime

54.39

speaker

1

averageConfidence

0.8265822222222222

speaker_name

Eric Glenn

index

7

match

fred

filterType

noteFilter

incoming

false

metrics

1

word

Fred

category

People

2

word

Barrick

category

Nouns

3

word

Fred

category

Nouns

sentiment

0

sentimentType

neutral

questionFilter

false

importance

0.6503729820251465

readability

0.718589723110199

task

0.9329748153686523

algScore

21.604229755609214

importanceScore

12.050266608413384

taskMatch

false

id

7

4

id

3

note

See if we can find a tutorial app here.

action

true

question

false

caption

sentence

See if we can find a tutorial app here.

speaker_id

1

time

104.08

endTime

107.07

speaker

1

averageConfidence

0.8418077777777778

speaker_name

Eric Glenn

index

11

match

we can><find

filterType

taskNoteFilter

incoming

false

metrics

sentiment

0

sentimentType

neutral

questionFilter

false

importance

0.6153966188430786

readability

0.8204665780067444

task

0.970687747001648

algScore

25.51338827767414

importanceScore

15.767366372621977

taskMatch

false

id

11

5

id

4

note

Choices text appears to be the common text.

action

false

question

false

caption

sentence

Choices text appears to be the common text.

speaker_id

4

time

204.36

endTime

207.22

speaker

4

averageConfidence

0.82783875

speaker_name

Eric Glenn

index

16

match

none

filterType

none

incoming

false

metrics

sentiment

0

sentimentType

neutral

questionFilter

false

importance

0.3249487280845642

readability

0.8004507422447205

task

0.7312167882919312

algScore

18.759122942739285

importanceScore

9.183560801106829

taskMatch

false

id

16

6

id

5

note

Okay, we've run it for four minutes.

action

false

question

false

caption

sentence

Okay, we've run it for four minutes.

speaker_id

6

time

209.8

endTime

211.972

speaker

6

averageConfidence

0.9484085714285715

speaker_name

Eric Glenn

index

18

match

none

filterType

none

incoming

false

metrics

1

word

four minutes

category

Numbered Nouns 1

2

word

for four minutes

category

Dates

3

word

four

category

Numbers

4

word

minutes

category

Nouns

sentiment

0

sentimentType

neutral

questionFilter

false

importance

0.2647213935852051

readability

0.7716603875160217

task

0.9792461395263672

algScore

19.60311531233475

importanceScore

10.07487112681071

taskMatch

false

id

18

customTopicSentencesWithContext

meetingTaskNotes

We have an automation test with Barrick and Fred.

See if we can find a tutorial app here.

meetingMetricNotes

Okay, we've run it for `four minutes` .

meetingPricingNotes

I have a prompt that features some test data, but it's not the correct.

Data that I was looking for.

We have an automation test with Barrick and Fred.

That is the same question.

Okay, we've run it for four minutes.

meetingQuestions

meetingDateTimeNotes

I have a prompt that features some test data, but it's not the correct.

Data that I was looking for.

We have an automation test with Barrick and Fred.

That is the same question.

Okay, we've run it for four minutes.

summary

• This test is designed to evaluate the automation system with Barrick and Fred. It involves running a tutorial app, which appears to be successful after four minutes of testing.

• Keywords: automation test, barrick, fred

• Share feedback: https://airtable.com/shrTJHhCq2PHdC3fl

Userlevel 7
Badge +14

@E_glenn 

We’d need to see a full list of the data points returned from the Zap trigger step.

 

 

 

This post has been edited by a moderator to remove personal information. Please remember that this is a public forum and to remove any sensitive information prior to posting.

When I checked the history of the last run I compared what the total transcript covered vs what zapier sent chat gpt. I relized that the full transcript was not being sent into chat gpt. It may have something to do with the inserted data in the prompt. I want to send the full transcript from Fireflies.ai into chat gpt. 

Userlevel 7
Badge +14

Hi @E_glenn 

Good question.

For clarity and context, please post detailed screenshots with how your Zap steps are configured along with the encountered error, thanks.