Sisense Intelligence token estimation

Hi.

We are considering activating Sisense intelligence to use a LLM for our data models and dashboards.

We want to estimate how much this would our organization from the LLM provider side. It depends on the tokens that are input into the model. In the Sisense documentation it mentions that depends on the number of tables and column of the model, but I guess also on the pre-instructions text that is sent along the user's input.

Is there a way of knowing how many tokens are added, or how much would cost to create a medium size dashboard given the OpenAI pricing?

Thanks,

Javi.

vatojavier

Posted 1 year ago·Last reply 1 year ago

4 comments

DRay

·1 year ago

yaellev1, Are you able to answer this question?

yaellev1

·1 year ago

vatojavier that is correct, both the user prompt and some system instructions are passed to the llm along with the datasource semantic metadata , contributing to the overall token count. We plan to introduce in product analytics to help customers understand their average token usage this is not available at this time. This is currently on our 6 months roadmap.
More information on what data is processed across different use cases can be found here:
https://docs.sisense.com/main/SisenseLinux/data-flow-per-feature.htm?tocpath=Sisense%20Intelligence%7CData%20Handling%20-%20Securing%20AI%20in%20Sisense%20Intelligence%7C_____4