Hi Team,
Earlier, we were able to train the AI efficiently, and the responses were accurate and reliable. Although AI token consumption was higher, but both our team and our clients were satisfied with the quality of the AI's responses.
Since the introduction of vector-based training to reduce AI token consumption, we have noticed a significant drop in response quality. The AI is no longer responding accurately, and it has become much more difficult to train it effectively. This has impacted multiple client implementations and has resulted in frequent issues with incorrect or repetitive responses.
We kindly request that you either:
- Restore the previous AI training method, or
- Provide a toggle option that allows us to choose between the previous training method and the new vector-based training.
This would allow us to continue delivering the level of AI performance our clients expect while giving users the flexibility to optimize for either response quality or token consumption.
This issue is affecting multiple clients, so we request that it be prioritized and addressed as soon as possible.
