Since the token we input to LLM is limited, the history messages we add into gpt request is also limited.
This merge request defines the interface MemoryMessagePicker, it can pick up needed chat messages from memory. We also provide an implementation of this interface BufferWindowMemoryMessagePicker, which will pick up lastest 2*K messages from memory.
Since the token we input to LLM is limited, the history messages we add into gpt request is also limited.
This merge request defines the interface
MemoryMessagePicker, it can pick up needed chat messages from memory. We also provide an implementation of this interfaceBufferWindowMemoryMessagePicker, which will pick up lastest 2*K messages from memory.