A0 Docekr Ports or LiteLLM Issues?

4d (edited) • ❓ Q&A

Hello and Good Afternoon Everyone,

Please forgive my wall of text. I will state that I'm not 100% familiar with docker, as I've only used it on my desktop and not in the wild. Also, when discussing technicals on AI products I do believe that "context" is important for transparency.

I have installed A0 on my Debian VPS. I followed the instructions found here -> https://www.agent-zero.ai/p/docs/get-started/#installation Docker components would not install as stated. Had to first add docker source and key, then I was able to install. Followed everything else; was able to pull image and start with persistance just fine.

For added layer of security I have installed Tailscale - private vpn from my home -> A0 host (for free).

I was able to get to the WebUI just fine by exposing port 50080 to tailscale (tailscale serve) which gives me an https URL only available on the tailscale network.

Now for my nftables firewall - I am only allowing all connections through tailscale inteface; everything else gets denied by policy. I can ssh in, ping out, etc. Secure two-way communitcation established at the host/network level.

After I installed A0 and able to login to the WebUI - I went to settings as instructed to set API keys. I found the area for Embedding Model. I set my model provider (openrouter), set my model (as suggested in documentation openai/text-embedding-3-small), set my API key, and set the embedding model base URL. Click "Save".

I go to open "new chat' and type 'hello' -- starts "Initializing memory database...". It continues to do this for some time. I look in the logs (docker logs --tail 100 -f agent-zero) and I see this:

WARNING [huggingface_hub.utils._http] Retrying in 8s [Retry 5/5].

'(MaxRetryError('HTTPSConnectionPool(host=\'huggingface.co\', port=443): Max retries exceeded with url: /sentence-transformers/all-MiniLM-L6-v2/resolve/main/tokenizer_config.json (Caused by NameResolutionError("HTTPSConnection(host=\'huggingface.co\', port=443): Failed to resolve \'huggingface.co\' ([Errno -3] Temporary failure in name resolution)"))'), '(Request ID: f16d5bf3-43e9-493a-803a-d30b16d0f987)')' thrown while requesting HEAD https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2/resolve/main/tokenizer_config.json

WARNING [huggingface_hub.utils._http] '(MaxRetryError('HTTPSConnectionPool(host=\'huggingface.co\', port=443): Max retries exceeded with url: /sentence-transformers/all-MiniLM-L6-v2/resolve/main/tokenizer_config.json (Caused by NameResolutionError("HTTPSConnection(host=\'huggingface.co\', port=443): Failed to resolve \'huggingface.co\' ([Errno -3] Temporary failure in name resolution)"))'), '(Request ID: f16d5bf3-43e9-493a-803a-d30b16d0f987)')' thrown while requesting HEAD https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2/resolve/main/tokenizer_config.json

This raises two questions for me:

Question 1: Do I need to expose addtional ports from docker to tailscale (port 80, port 443, etc)?

example: (Caused by NameResolutionError("HTTPSConnection(host=\'huggingface.co\', port=443): Failed to resolve \'huggingface.co\' ([Errno -3] Temporary failure in name resolution)"))')

Question 2: Why is A0 going to Hugging Face? I think I read that by default A0 uses "/sentence-transformers/all-MiniLM-L6-v2" for memory. But I set Openrouter w/API key and URL in the settings, why is it still trying to go to HuggingFace?

Thank you for you time and patience,

~Martin

3 comments