Curl to HTTP need help in sending to localai.io install of whisper model

Help me Build my Workflow
1

Hi,
I have a local install of localAI and can transcribe mp3 audio to text using the below command

curl http://localip:8082/v1/audio/transcriptions -H “Content-Type: multipart/form-data” -F file=“@/data/shared/abc.mp3” -F model=“whisper-1”

I get the needed output when I run this from command line.

Now I want to leverage n8n local install running out of docker container and came up with the below template.

image
image814×401 13.6 KB

image
image790×330 13.7 KB

I keep getting HTTP Error 404 or 500 etc

I tried both POST and GET and cannot get this to work. Any suggestions is appreciated.

I am running n8n using docker desktop on windows 10 host. Mapped local folder to /data/shared on the container. When the local folder on host is added with a file, new file event is triggered and this file is sent to localAI setup using n8n_HTTP

Please let me know if any additional info is needed.

2

Couple of things to try:

  1. From your pasted workflow, the input data field name in http request node is empty. Check the output from the file node to see what the binary field is called, and add that to the input data field name.
  2. Since n8n is running in docker, you might need to use http://host.docker.internal instead of http://localhost

If that doesn’t help can you share the http node input and output view of one of the failed workflow runs

3

Hi Richard,
Thank you for your response.

Since I was not making progress using n8n, I moved to CrewAI and was able to make significant progress.

I understand n8n and CrewAI provides different capabilities.

Also HTTP request was not working at all (even with CrewAI) and using curl with a defined syntax was mandatory to get the desired result of transcribing using whisper-1 model running locally.

As of this moment I think HTTP cannot be used for my use case.

Once again thank you for taking time to respond.

With Best Regards,
Subba