Currently the agent is over using the calculation capability, this behavior can be broken into different steps where the LLM first plans how to write useful calcs and then formulates a query that uses said calcs.
Breaking this apart will allow for more detailed prompts used in the calcs and query steps of the chain
Currently the agent is over using the calculation capability, this behavior can be broken into different steps where the LLM first plans how to write useful calcs and then formulates a query that uses said calcs.
Breaking this apart will allow for more detailed prompts used in the calcs and query steps of the chain