Unverified50% confidenceFactExact time
DUMEdb directly audits the transcript.jsonl file on disk, computes the real token consumption after compaction, and triggers a clean session handoff before automatic compaction causes amnesia.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/20/2026
First Seen
Valid until: 1/18/2027
Sources
Related Entities
Related Claims
UnverifiedGBrain uses plain-text Markdown files to store memory, sacrificing some semantic retrieval precision in exchange for complete auditability and portability.70% similarUnverifiedMemory compression retains semantic cores by summarizing and distilling conversation history, but the compression process may lose structured constraint information.70% similarUnverifiedSummary Buffer Memory is a hybrid of Buffer Window Memory and Summary Memory, retaining the full content of the most recent k messages while compressing the excess into a summary.69% similarUnverifiedSelective memory compression uses the decision model to determine whether a certain execution segment is 'closed'; if closed, it is collapsed into a semantic summary, otherwise the original log is retained.69% similarUnverifiedSummary Memory compresses old messages into a summary before they are deleted, and the summary is usually generated with a cheap small model to control cost; in LangChain this corresponds to SummarizationMiddleware().67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1075035API
curl https://kongchang.com/api/v1/knowledge/claims/1075035MCP
get_claim(id=1075035)