Unverified50% confidenceFactExact time
In v0.0.46, auto-compaction uses the context window of the selected model, and falls back correctly even if it is not provided in the directory entry.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/13/2026
First Seen
Valid until: 1/11/2027
Sources
Related Entities
Related Claims
UnverifiedHeadroom's compression is reversible; compressed content sent to the model is cached on the local machine and can be retrieved at any time.70% similarUnverifiedCommon token compression strategies include pre-filtering retrieved content, limiting context window size, and consolidating multiple small calls into a single structured call.60% similarUnverifiedParameter compression refers to relay services reducing a model's context window or output token limit, such as compressing Claude 3.5 Sonnet's 200K context window down to 32K or shorter.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1019198API
curl https://kongchang.com/api/v1/knowledge/claims/1019198MCP
get_claim(id=1019198)