Is it possible to access the Ollama API

[dadofsambonzuki]

There are some VSCode plugins that can access local LLMs. I'm using Continue.

Is it possible to access the local Ollama / Alpaca API so that extensions can use?

[Jeffser]

Hi yes I just checked and it works!

What you gotta do is go to ~/.var/app/com.jeffser.Alpaca/config/server.json and change local_port to 11434

Then on the Continue menu go to Local with Ollama if it doesn't connect automatically just click the Ollama URL and it will try to connect itself

If you have any other questions feel free to ask!

[dadofsambonzuki]

Nice! When I change the port number to 11434, it crashes after a few seconds. Changing back (or to another value) fixes it again.

I must have a port clash from a borked install/removal of Ollama, although I don't see anything running on 11434.

EDIT - It does briefly start-up on 11434 before bailing.

[Jeffser]

yeah there's probably something else using that port, that's why the default one is different

[dadofsambonzuki]

I've got ...

127.0.0.1:11435  users:(("ollama",pid=6482,fd=3))                                                       
127.0.0.1:44203  users:(("ollama_llama_se",pid=5350,fd=3)) 

The second service continues to run after I close Alpaca. Nothing on 11424 according to ss.

[Jeffser]

Just kill those processes and try again, sometimes Ollama keeps running when it isn't supposed to

[dadofsambonzuki]

5350 killed.

No change to starting-up on 11434 - i.e. it starts, then aborts.

Nothing obvious in tmp.log.

What I have found is if I try and start a second time it does load.

[ntnsndr]

Exciting to see that this works. Has anyone tried it with the emacs chatbots, like Ellama or gptel? I have just started tinkering and haven't gotten either working, but I haven't tried all the configuration options yet.