[BUG] Gemini Pro - NO API Configuration

[michsiu]

Found a bug? Please fill out the sections below. 👍

There should be a API configuration for Gemini Pro "apis" file

Describe the bug

On the official doc, there should be a API configuration for Gemini Pro before run Gemini, like this

GOOGLE_API_KEY=userdata.get('GOOGLE_API_KEY')
genai.configure(api_key=GOOGLE_API_KEY)

But there is no this step on the self-operating-computer

After added the API configuration, it works now

Another Bug About API Configuration

As I searched on google generativeai issue on GitHUb, it notes that it would be better add transport='rest' on API configuration of Gemini Pro
So, the correct code should be
genai.configure(api_key=YOUR_GOOGLE_API_KEY,transport='rest')

[joshbickett]

@michsiu thank you for reporting. I'll take a closer look

[sangeet000007]

can it run completely without openai api key?
i added
GOOGLE_API_KEY=userdata.get('GOOGLE_API_KEY')
genai.configure(api_key=GOOGLE_API_KEY,transport='rest')
but its still asked to set open ai key after setting null api key it enter into the software but returns "Gemini Invalid model"
should i add authorize credentials for a desktop application and i followed all the steps and got the json file but i dont know where to set those credentials and there is no tutorial in readme so kindly any one provide proper tutorial to work with gemini only.

[joshbickett]

@michsiu I can confirm that this .configure appears to fix the issue. I am working on a branch to fix this and will merged once ready

[joshbickett]

@sangeet000007 you should be able to run the gemini-pro-vision model without an OpenAI key saved. You'll only need a Google key. I just fixed some issues with the config of these keys. I think it should be working correctly. Let me know if you have any issues

[joshbickett]

@michsiu I added genai.configure(api_key=GOOGLE_API_KEY) as you suggested and it appears that gemini-pro-vision is now working as expected. Could you try the newest version and let me know if it works for you?

[sangeet000007]

its working
the click of button is not quite good with gemini it cant even send mails i suggest to improve gemini models with techniques like Set-of-Mark Prompting gpt-4 and also give support for open source models
Thank you.

[michsiu]

Thanks @joshbickett ,it works