LLM Flow is a Large Language Model Work Flow Framework.
The fundamental construction for the LLM Flow is a direct graph (no need for the graph to be a DAG), in Computer Science.
A node is something that receives an input: a string(In some particular case, the node can take nothing) and returns an output: also a string(In some particular case, the node can give nothing), will be either passed to another node, or directly shows to user.
There is some pre-defined node types to make development easier.
Edge is consider to be the pass process mentioned in the Node part. Which means that neither the developer nor the user need to consider about it.
This is the base class of every node type, it is more like a interface, though it can return what it receives.
This is the base class of every node type that needs to interact with LLM.
An Inputter takes the user input(for developer, it takes nothing) and directly returns it.
An Outputter takes a string and directly shows it to user(for developer, it returns nothing).
An Extractor takes a string, extracts informations out of it, structures it and returns it.