search

crazyyanchao / llmcompiler

LLMCompiler is an Agent Architecture designed to speed up the execution of agent tasks by executing them quickly in the DAG. It also saves the cost of redundant token use by reducing the number of calls to the LLM.
Apache License 2.0
18 stars 3 forks source link

A use case example #2

Open nooralahzadeh opened 3 months ago

nooralahzadeh commented 3 months ago

Could you provide an example of show case of how your LLMcompiler provides a plan and final answer? Thanks

crazyyanchao commented 3 months ago

@nooralahzadeh Please refer to the example below. Example of Performing Complex Mathematical Calculations

nooralahzadeh commented 3 months ago

Thanks @crazyyanchao. Is there any way that we use the StateGraph instead of MessageGraph and update the state in each task?

crazyyanchao commented 3 months ago

@nooralahzadeh I would consider implementing a stategraph of state updates for the LLMCompiler. Also, can you add a specific reason why you'd like to use stategraph?

nooralahzadeh commented 3 months ago

@crazyyanchao In specific use cases, you need to pass a JSON or data frame structure from one step to another step and add or complete it during the steps in the graph. Having a state graph will help a lot!