ferhimedamine commented on issue #384: URL: https://github.com/apache/flink-agents/issues/384#issuecomment-4734921869
Batched summarization for long-term memory is a solid approach — the question is what to summarize and at what granularity. We found that combining summarization with importance scoring gives better results: high-importance memories stay verbatim while low-importance ones get summarized into knowledge graph entities. The entity extraction acts as a natural summarization layer — "Sarah from Engineering decided to use Redis" becomes a graph edge (Sarah decided Redis, context: Engineering) that's cheaper to store and faster to traverse than the full text. Full implementation: https://github.com/Dakera-AI/dakera-deploy#used-for -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
