dsh-compaction-micro
Tiered micro-clear layer for DeepSeek Harness compaction: clears whitelisted re-runnable tool results to placeholders before the official summarization tier, freeing context with no model call.
840 results
Tiered micro-clear layer for DeepSeek Harness compaction: clears whitelisted re-runnable tool results to placeholders before the official summarization tier, freeing context with no model call.
DSH Web plugin: a right-docked file panel that browses any directory and views/edits text, images (brightness/contrast/grayscale/rotate/flip), and Office/PDF documents (.docx/.xlsx/.pptx/.pdf via local Python).
88API Image Studio for DSH: six Image2, Image 2.5, and Nano Banana models for text-to-image, multi-reference editing, 2K/4K output, and sequential batches.
Agnes AI image and video generation for DeepSeek Harness — agnes_image and agnes_video tools over the Agnes OpenAI-compatible gateway.
Out-of-tree dsh plugin: GitLab, GitHub, Gitee, Gitea, and Bitbucket tokens stay out of the model context, stored encrypted (AES-256-GCM) in a plugin-owned file; the model calls forge API tools on demand, and the web settings page manages sites and tokens.
Conversation-local credential handoff that writes through the DSH credential service without returning the secret to model context.
Curate tool output before it enters the model's context: keep error/warning lines and head/tail, spill the full text to disk with a pointer. Faster turns, smaller context, sharper attention — for every tool, not just bash.
DeepSeek Harness 上下文自动压缩、手动压缩与溢出恢复插件
最新文档注入防幻觉:编写/引用库代码前先拉取版本化最新文档(ctx7 CLI 或 MCP),library ID 直取、版本指定、规则自动触发;消灭过期示例与幻觉 API。受 upstash/context7(61k★ MIT)启发。
MCP server studio for DeepSeek Harness: live-mounted servers with real connection status, a per-server tool browser, JSON import, and a dashboard-grade settings page.
Quick prompts dock for the dsh web GUI: a row of user-defined prompt chips above the composer — click a chip to preview and edit its text (with {{placeholder}} fields), then sync it into the input or send it directly. Prompts are stored in the official settings document, not localStorage.
AntV AVA and GPT-Vis plugin for DeepSeek Harness: stream data-heavy answers as narrative text, charts, or both.
A refutation ledger for coding agents: remember what did not work, and expire it automatically when the code it depended on changes.
A DeepSeek Harness plugin that integrates claude-mem: query, context injection, manual save, and session lifecycle over a local claude-mem worker's HTTP API
Vision for text-only DeepSeek Harness agents: delegate image reading to a one-shot subagent on a configurable vision route (MiniMax / Kimi / any OpenAI-compatible provider), keeping image bytes and the vision model's context out of the main session.
DSH multimodal input plugin: route file attachments (images, video, audio, text) through per-preset model chains and feed the results to the session model as prompt tokens. Adds a Multimodal settings page.
mm-vision (通感编码器) for DeepSeek Harness — give any text-only LLM the ability to see images via structured spatial text encoding. Registers the mm_vision tool.
Enforced pain-point-check guard plugin for DeepSeek Harness: after two non-converged experiments on one problem it injects the three questions, denies non-investigative tool calls until they are answered in the reply text, and blocks same-direction retries
DSH 悬浮 API 余额徽章:可拖动、可缩放、亚克力质感、文字随背景明暗自适应。支持 DeepSeek / Moonshot / OpenAI / 自定义接口。· A floating API-balance badge for DeepSeek Harness: draggable, resizable, acrylic, adaptive text color.
Local-first vision for text-only DeepSeek Harness agents: route image understanding to a local OpenAI-compatible vision model, return structured JSON evidence (summary / OCR / layout / semantics / visual / uncertainty).
Gives a text-only LLM vision capability
DSH plugin: vision_read — route image reading to a dedicated vision model (e.g. Kimi K3) so text-only agents can see images
Agentic Surface Compaction (ASC) for DeepSeek Harness: the model decides when and what to compact, committed as durable session-log replacements with full replay, search, and degradation
DSH Web 语音输入插件:默认使用浏览器内置 Web Speech API,也可按需下载本地模型离线识别。