The "context-compaction" topic on GitHub groups 14 open-source projects in the DeepSeek Harness (DSH) ecosystem, led by dsh-force-compact with 10 GitHub stars. dsh-force-compact — Aggressive context compaction for local-first agents. Every project here is indexed by DSH Universe with live GitHub data — stars, activity and install status — so you can compare and install directly.
Aggressive context compaction for local-first agents. Runs Qwen3.8‑27B on self‑hosted llama.cpp at low context, shrinking history so the live prompt stays small, fast, and private—delivering a big‑window experience without API cost or data egress. Aggressive context compaction plugin for local-first setups. Runs Qwen3.8‑27B on self-hosted llama.cpp at low context, constantly shrinking history to keep the resident prompt small and fast, combining privacy with a big-window experience, zero API cost, data never leaves the machine.
五种载体,一座本地对话档案馆:原文归档、分层后台摘要、原文查证与跨工具接续。默认原生压缩,Claude Code 与 dsh harness 可选接管。Local archive, layered summaries and exact-source recall across five harnesses; native compaction by default.
Adds automatic context compaction, /compact, /context and model-initiated compaction to DeepSeek Harness minimal mode, fixing the problem that long tasks cannot keep running in minimal mode.
DSH 会话越长越慢、每轮越贵?把最老的历史折叠成一条带索引的消息,全文归档到本地、随时能沿路径找回来,不调用模型写摘要。DeepSeek Harness plugin: folds the oldest history span into one indexed message backed by a lossless local archive.
Native DeepSeek Harness (DSH) plugin: reversible incremental context compaction with model-facing tools (status/search/restore/compact) and a Web panel — save tokens on long sessions without dumbing the model down.