AI Integration

[random09456]

I can imagine that AI is a controversial topic. But I would actually appreciate it if the entries were explained in a bit more detail.

Maybe that could be added even without AI?

[Belim]

a local description database is pretty much out of the question 😄
i really dont feel like manually describing 4000+ Winapp2 entries, Ccleaner never did that either

i might enable the optional AI explanation feature for everyone in the next Insider build though

[KniggeMS]

The AI option is really great!

[random09456]

Totally understandable 💯
As an optional feature, it really should be fine. I hate AI slop too, but an AI-generated explanation - provided it can be disabled -seems like a good compromise.

[Belim]

feel free to test it in the current Insider build
https://github.com/builtbybel/FluentCleaner/releases/download/26.05.02/FluentCleaner-Insider-win-x64.zip

happy to discuss feedback here or on X as well
https://x.com/builtbybel/status/2053473663765102665

[random09456]

Thanks! Good implementation

[random09456]

hank you for the further development.

Would it also be possible to implement the Google Gemini API? You can get an API key for free here as well.

[Belim]

technically yes but one reason i started with Groq is that the setup process is way less painful for normal users
Gemini is free too, but google cloud tends to immediately turn into create project >> enable APIs >>billing dashboard PTSD 🥲

[xzhlksh]

It would be even better if an OpenAI-compatible API could be used. I have plenty of code plan, but I’m unable to register for Groq to use them.

[spmet-apps]

I thought there was a bug but it looks like OpenAI keys are not supported (yet) so I second @xzhlksh --- when possible it would be nice to be able to use it since some of us are already setup in OpenAI for multiple projects.

[imtipi]

same,OpenAI-compatible API is better,so that i can use other opensource models for ai explain

[Belim]

@imtipi @spmet-apps @xzhlksh do you mean a custom openai compatible endpoint for services like ollama, lm studio, openrouter, or a self-hosted model? fluentcleaner already supports groq, openai, and anthropic directly, so i just want to make sure this is what youre looking for #115

[imtipi]

yes,custom openai compatible endpoint is what i need,because i am using thirdprovider like openrouter.
maybe you can just allow custom endpoint and model name as advance option when user select openai, so that i can use openrouter(or other provider) and you can just reuse the exist code(i think?)

[Belim]

i actually enabled it in the current Insider build as an experiment ( ive been playing around with custom OpenAI-compatible endpoints for a while now, but i'm still not sure whether it'll make it into the stable release

feel free to give it a try and let me know how it works with OpenRouter
https://github.com/builtbybel/FluentCleaner/releases/tag/classic-1.10.091
check assets

for now it's only available in the Classic edition. the Modern/WinUI version is still quite a long way from getting this

[imtipi]

works for me

also tried other provider and works too

although after i can use the ai feature,i just realized it can explain static Winapp2 entries only,i was though it can explain to me according to what files i scanned so that ai can actually help me confirm that it is safe to delete.but myabe i am going too far because ai might have hallucination and a single time of hallucination might cause user lost valuable files...

i really dont feel like manually describing 4000+ Winapp2 entries
generate the inital database by using ai still possible,but it cost maintainer lots of time of correction pr(no matter submit by maintainer or regular user) in the future,maybe let us put this idea aside

anyways,custom OpenAI-compatible endpoints works for me,thanks

[Belim]

glad to hear the custom endpoint works

and yes... it can go a lot further than what's currently public. i have a private build that can reason about the actual scan results (i've been experimenting with Kimi K3 as well), not just static Winapp2 entries

but i'm honestly not convinced it belongs in a public release. one confident hallucination is all it takes to hurt the project's reputation, and i'm not a big fan of pretending AI always knows best
so for now i'm keeping that part on the experimental side

[Chaprnks]

I'd really like to see support for inputting your own Gemini API key. They always seem to be the last option for developers to support, ironically over OpenAI

[random09456]

I would also appreciate it if Gemini API were supported

[IAmSympathy]

I'm using Groq. Fluent Cleaner (Modern) says the API key is invalid, but the same key works when I manually call the Groq API from PowerShell (e.g. GET /openai/v1/models).

[sx360to-maker]

Works with local ai models, Gemma4 doesn't seem to work properly, Tested with qwen3-vl-4b and it works fine, I prefer the modern Fluent cleaner, When will it have Open AI Providers?