Solidify DSH session history step by step: rewrite tool results in place to the facts the model actually kept, so long conversations stop re-sending noise every turn. Costs tokens and time: a ~770-character contract in every request, plus one extra full-context request and round-trip per step when stepSummary is on.
Independent dsh plugins for things upstream can't (yet) accept as PRs. Starts with @zynieie/dsh-lan-fix: let dsh web load on http://<lan-ip>:3080 (insecure context) without the WebSocket abort storm.
Persistent, model-editable memory/notes store for DeepSeek Harness. Adds memory_set/get/delete/search tools backed by ctx.storageDomain so facts survive across sessions (dsh-plugin).
Multi-instance ComfyUI orchestration for DeepSeek Harness: give the agent a live view of every GPU in the farm and dispatch each generation job to the idlest instance instead of always hitting the busiest endpoint.
为 DSH 提供持久化输入历史,支持使用键盘上下方向键切换历史记录并恢复未发送草稿。Persistent input history for DSH with Arrow Up and Arrow Down navigation, draft recovery, deduplication, and local storage.
Mr. Meeseeks: heterogeneous free-model teams for DeepSeek Harness (dsh-agent-teams replica with Meeseeks art) + Budget-AGI brain-and-swarm preset + swarm scripts
Cross-session user-profile memory for DeepSeek Harness: remembers who you are and how much you know, so the agent speaks at your level. Approval-gated, auditable, local SQLite.
DeepSeek Harness plugin: when a run finishes or needs approval, get a link on your phone, hear the result from an ElevenLabs voice agent and say what to do next
DSH token-saving toolkit: proactively slim large text/JSON/CSV before they enter context (diet_text / diet_json / diet_csv / diet_estimate). Zero-dependency pure functions.