Constant restarting of "advanced workers" container because missing DEEPGRAM_API_KEY #159
Description
Current Behavior
The advanced-workers container is constantly restarting:
Now the funny side effect is that you get anything from 30 - 150 redis workers ready for work 🤖🤖🤖
Expected Behavior
The advanced-workers container should be setup in a healthy way and only 8 redis workers should be setup like the start-worker.sh implies.
Possible Solution
I think the problem is, that there is no parakeet support right now in main.
That is why there is only one error in the docker logs of the advaned worker container -> the missing deepgram api key.
Steps to Reproduce / Context (Environment)
I used the wizard.py for the inital setup and the following options:
My plan is to use the omi backend only offline for privacy reasons. I have an ollama service running raw on the hardware and using docker compose for everything else.
Possible Implementation
I think this PR already fixes the problem as far as the missing parakeet integration.
#150
Other fixes / problems i had during the plain start
Missing connection to speaker service
The friend lite backend did show that there is no connection to the speaker service. I do not know why it could not use the normal host.docker.internal URL but if you change the url to the docker DNS name for the speaker service container, it will connect.
So this is from my .env file in the advanced friend backend directory.
SSL Internal Error when connecting over tailscale
I tried debugging this with OpenSSL and curl. It seems that Caddy doesn’t allow using an IP address as an alternative DNS name. Regular browsers and curl don’t send the IP address as the domain name during the TLS handshake, but Caddy expects them to do so. A simple fix is not to put in the tailscale ip as DNS name as the wizard recommend but using the "magicDNS" name you get from your tailscale account for this server.
This is the same .env file as above. Or if you setting it via the wizard.py:
