Open KTheMan opened 3 weeks ago
So maybe a clearly defined interace to an external model could solve this.
I don't know if we can bulild this as an "easy to use for everyone" solution but could provide an interface which connects to some external model via a few lines of code.
Yup, pretty much what I was thinking; shouldn't need much, as Ollama exposes an OpenAI compatible completion endpoint.
From https://ollama.com/blog/openai-compatibility:
import OpenAI from 'openai'
const openai = new OpenAI({
baseURL: 'http://localhost:11434/v1/',
// required but ignored
apiKey: 'ollama',
})
const chatCompletion = await openai.chat.completions.create({
messages: [{ role: 'user', content: 'Say this is a test' }],
model: 'llama3',
})
I did a bit of the work over at https://github.com/KTheMan/UnKenny. I haven't gotten far enough to get a proper response, but it does hit the Ollama endpoint
Ah I see. Going for a fork right now was also what I would have suggested. I should be enough to change openai-api, as you did. What is the error you get? Or how does your response look like?
I have Ollama instance set up on an external server and think it would be a great, easy fit for this type of add-on if I could hook it in. Opens up use of many more models with better hardware, if the GM has access, IMO