autocompletion does not work in VSCode

[arturshevchenko]

Describe the bug

autocompletion does not work at all

How to reproduce

install olama, run ollama, pull deepseek-coder, run deepseek-coder
write in configs deepseek-coder

write some code and wait untill completion

Expected behavior

code completion works

Actual behavior

autocompletion does not work

[srikanth235]

Hey,

Could you please check the logs by using the 'Privy: Show Logs' option in the command palette? Check to see if the requests are being sent to the Ollama instance. You should see something similar to the attached screenshot.

89/ce035e4b-0095-4c74-8607-e6b92ab2e971">

[arturshevchenko]

in logs response is shown

[srikanth235]

For auto-completion, can you please try with base models as shown in the image.

[arturshevchenko]

the same :(

[srikanth235]

I wasn't able to reproduce this issue. A few things to check are:

1. Are there any other installed extensions that also provide autocompletion. If yes, please try disabling them.
2. Please try updating your VSCode to the latest version.
3. Try re-installing the extension and restarting the VSCode editor.

[rainabba]

@srikanth235 Thanks for the tip about using 1.3b-base. I too was trying to use a non -base model.

@arturshevchenko, from his tip I realized the naming held significance here and the fact that I was able to use the 7.8b-base, I think the "-base" indicates training that scopes the Q/A to code. You got a response, but it was verbose and didn't have the <> tags I'm seeing in my now-working query/response. If your issue is the same, switch to one of the 3 "-base" models depending on your hardware capabilities. For some reason, ollama isn't using my GPU now though so responses take so long it "look" like nothing is happening until I watch the Privy log output.

@srikanth235 Thanks for the awesome extension!