Idea: long-term memory for the agent across restarts

#17
by Corbenic - opened

Your official Flowise Space is a clean visual way to build agents โ€” a ChatOllama node can go into a Tool Agent or a Conversational Agent, with retrievers and other tools attached. That is a real local path, but if Ollama or the Space restart, the agent must recompute the chat memory, the tool schemas and the system prompt again. Galahad is long-term memory for the agent, it can save and restore the KV cache byte-for-byte across restarts so the agent skip recomputation and keep memory between sessions. We also see 8.8 to 12.3x less GPU energy for that step, and faster first token on all 7 NVIDIA GPU types tested. Would you be open to try it, or do you want some help to integrate? https://github.com/corbenicai/galahad

Sietse

Sign up or log in to comment