"An error occured when attempting to chat with LLM" error message

[VE1GA]

I'm trying to set up the extension for the first time using Gemini, I have Google AI Pro for Students (I got it for free using my university email, though I’m not sure if that affects something).

• I followed the YouTube tutorial and tried multiple models but I keep getting an error message.
• I enabled the Debug option for more details, but I'm still unable to identify the fix (see attached settings and error logs).
• My goal is to use the models available to me for free (if possible).
• I'm using LLM Extension v0.0.2.0 and PowerToys v0.98.1 (v0.9.10852.0 Command Palette)

---

Could this be related to the AI Studio Status, and not a mistake on my side?

[LioQing]

I couldn't call that model as well, I think gemini-flash-lite-latest is not a valid model? check https://ai.google.dev/gemini-api/docs/models for a list of models from google ai studio. the closest model name i could find is gemini-flash-latest which is already deprecated

[VE1GA]

Today I tried using the same configuration shown in the screenshot. It replied to four test messages, but now when I press Enter, the messages get sent and a loading animation appears, but then nothing happens. I receive neither an answer nor an error.

I’m also experiencing an issue mentioned in #2 and #7: if I type too fast, the last few characters get deleted while I'm typing.
From what I’ve seen though, this should be fixed in a future update.

---

I couldn't call that model as well, I think gemini-flash-lite-latest is not a valid model? check https://ai.google.dev/gemini-api/docs/models for a list of models from google ai studio. the closest model name i could find is gemini-flash-latest which is already deprecated

Regardless, I switched to gemini-2.5-flash-lite to use a stable model and avoid future complications.

[LioQing]

I am aware of the deleting character issue, it was caused by the old command palette extension dependency, I'm working on publishing a new version of this extension with updated dependency soon, which should fix the issue

the issue with the stuck in loading animation is also possibly due to the older SemanticKernel dependency responsible for the API call, should also be fixed after the dependency update

[LioQing]

Just tried again today with gemini-flash-lite-latest in v0.0.3.0, it seems to work now 🤔 though I'd still avoid using models not listed on their doc