Financial damages caused by received errors $$$$

[karloti]

I used this library (generative-ai-android) v0.6.0 on Gemini 1.5 Pro and it caused me financial damage!

I was getting errors instead of replies from Gemini and ended up getting charged.
Something went wrong while trying to deserialize a response from the server. for: (Ask Gemini)
Caused by: kotlinx.serialization.json.internal.JsonDecodingException: Unexpected 'null' literal when non-nullable string was expected (Ask Gemini)

200 prompts cycled through, which I had to repeat every few seconds until I got a result other than an error.

So I was charged millions of tokens <128k.

Since Gemini 1.5 Pro is still in preview, and the Android library itself is not stable, who should bear the brunt, when I have not received other than errors.

How at such an early stage, something that is still a prototype can cause financial damage to my company?

We've wasted enough time testing, and now money.

I stopped the Gemini API because I don't plan on using the library until this issue is resolved.

I'm sure I won't be the only one, but I might be the first to suffer.

These mistakes are paid with real money.

Please tell me in this situation, which I have provided you with the most detailed information by email, who will take responsibility for the recovery of damages?

Actual vs expected behavior:

At first, I was very happy that there is already such a library, although the problem of storing the API key is not solved. Such a library is vital for the next generation of Android applications that need to quickly and conveniently implement generative AI solutions. I used the library for testing, and instead of contributing to its improvement, I managed to waste money and a lot of time. My expectations were not met. I expect that such decisions should be given some credit so that one can realistically assess how they are charged. Until May 30th, I didn't have any token counting accuracy, and now suddenly, in addition to them, the bills came in the billing. I think such incidents should be taken into account because there will be many people like me.

Any other information you'd like to share?

No response

[davidmotson]

marking as a duplicate of #160