#@wompfox any reason not to set up
1 messages ยท Page 1 of 1 (latest)
I punted on adding the telemetry bit but it does need to get added. For streaming, from what I could gather from the google genai client, the startChat interface doesn't support it. There is a generative streaming function but I don't think it's compatible with how we're doing things
I see a SendMessageStream which seems like what I want
oh yep sorry, that's exactly what you want. So SendMessage can be swapped for that
I was thinking of this one which wouldn't work https://pkg.go.dev/github.com/google/generative-ai-go@v0.19.0/genai#GenerativeModel.GenerateContentStream
sweeeet, will pick it up, hopefully doesnt take long
@dull dove pushed if you wanna try, added token metrics too
i switched back after it refused to do things i wanted it to ๐ https://v3.dagger.cloud/dagger/traces/b6761c6bf64b267543920532d5689223
LOL nice I will pull. Is this with gemini-2.0-flash?
no idea - might not be, I just accepted all the defaults in the AI studio and didn't specify on the CLI. I just set a token
cool I think that's default in our client at least if you didn't specify a model. I've found it's pretty dumb when we just give it a container or directory and ask it to use it, but really really smart with actual prompting and large contexts. Not sure why