deepseek-harness-compaction-ui
Visual percentage and token settings for official DeepSeek Harness compaction
399 results
Visual percentage and token settings for official DeepSeek Harness compaction
Evidence-backed working memory for DeepSeek Harness: a cited ledger of what already worked in this workspace, built from the session log you already have
Zero-dependency MCP stdio server template with an out-of-the-box A-share (Chinese stock) data toolkit. Key-free, NDJSON protocol, drop-in dsh bundle.
Fold each DSH Web turn's thinking, tool calls and context into compact expandable process boxes cut at every piece of model text, and reuse the shipped turn-process disclosure row as the toggle whenever it exists.
Token usage dashboard for DeepSeek Harness: cache, context, efficiency, per-step details, and cross-session comparison
DeepSeek Harness bundle: inject the user's IDE context (open editors + current selection) into each turn, via the Claude Code IDE WebSocket bridge (IntelliJ IDEA & VS Code).
Token-efficient tool-result pruning policy for DeepSeek Harness presets.
MCP adapter for DeepSeek Harness — one proxy tool instead of dumping every MCP schema into context.
把 git 状态与仓库规范动态注入 system prompt:分支/脏文件数/最近提交 + 仓库规范指引(官方 system-prompt section/context/variable 缝隙插件)
Deterministic privacy-aware repository context packs for DeepSeek Harness
DSH plugin: connect to MCP servers, discover tools, and register them for direct model invocation with retry logic and statistics
Context Transparency Manager: the model context as a visible, editable, effectiveness-scored surface (self-contained dsh bundle plugin)
Web sidebar indicator: API key context headroom, account balance, and estimated turns until compaction
MCP 服务器可视化配置插件:在 DSH Web 设置界面管理 MCP 服务器(stdio / streamable-http / SSE),保存即生效,无需手改配置文件。| Visual MCP server management for DeepSeek Harness: configure stdio / HTTP / SSE servers from the settings UI, hot-synced without editing config files.
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.
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.
@-mention context references for DeepSeek Harness: workspace file search and cross-session references from the web composer.
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。
Context compression for small-context models on the DeepSeek Harness Web GUI: one click condenses the whole conversation into a few sentences, opens a fresh session in the same workspace, and re-injects the summary as the first message so a small model keeps working with a clean, compact context.
上下文工程:按层级(规则文件→规格→源码→错误输出→对话历史)主动策划 agent 看到的信息,提升输出质量。受 addyosmani/agent-skills(88k★ MIT)启发。
dsh-docker — typed, guarded container control for DSH: structured docker/compose tools, project-aware targeting, an approval gate for destructive ops, service-health context, and a replayable status-table renderer
Load MCP servers for the DeepSeek Harness from the mcp.json of Claude Code, Kiro, OpenCode, and Codex, with user and project layers and live reload
LLM pre-review for sandbox-escalation approvals: an independent-context LLM gate answers sandbox escalation requests before they reach the user, falling back to the user on any failure.
Vision-first MCP for text-only Agents, using Agnes AI for image understanding with experimental generation and editing.