Skip to content

You need to remove nextSpeakerChecker or it will eat your brains #67

Description

@acoliver

What happened?

Google had a horrible design where Gemini-Pro is talking then Gemini-Flash decides if the conversation should continue. It sends the entire conversation multiple times per request.

It should be removed. (BTW it is still using the Gemini endpoint so all your qwen calls also go to google)

What did you expect to happen?

Not send the entire conversation mutliple times per request to google.

If you like, you can look at how I removed it here: https://github.com/acoliver/llxprt-code - presently, I only enable it for pro

(I also forked gemini and added multiple model support, including local.). You might consider doing your next version as a fork of llxprt and just remove the not-qwen support. I'd be pleased to work with you.

Client information

Details
$ gemini /about
# paste output here

Login information

No response

Anything else we need to know?

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    category/coreCore engine and logicpriority/P1High priority issuescope/model-switchingModel selection and switchingstatus/need-informationMore information is needed to resolve this issue.type/bugSomething isn't working as expected

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions