34 results
Eyes for text-only DeepSeek on DeepSeek Harness: image understanding via Doubao Web by default (zero cost, no API key), Antigravity IDE quota (flash/pro), Gemini API, or Cockpit proxy — model-invokable vision tool, wrapper adapters, evidence memory, and a bilingual client panel.
Vertical prompt quick-jump rail for the DeepSeek Harness web conversation view
One memory layer for every AI tool and agent, packaged for DeepSeek Harness with startup context, prompt-time recall, Mem MCP tools, and DSH thread capture.
AI 系统的数据层：把 18 种来源（文档站/GitHub/PDF/视频/Notebook/Wiki…）转成结构化知识资产，导出到 22 种目标（Claude/Gemini/OpenAI/RAG 管道/编码助手）；AI 驱动项目扫描生成配置。受 yusufkaraaslan/Skill_Seekers（14k★ MIT）启发。
网页复刻：从参考图/URL 提取字体、颜色、动画、组件结构，生成还原度高的网页项目。受 xuanxuan-prompts（169★）启发。
模型路由网关：一个端点连接 290+ AI 提供商（90+ 免费、1200+ 模型），Claude Code/Codex/Cursor/Cline/Copilot 全兼容——配额感知自动回退、19 种路由策略、RTK+Caveman 叠加压缩省 15-95% token、免费层聚合预算（~1.5B tokens/月）、MCP/A2A、本地优先。受 diegosouzapw/OmniRoute（51k★ MIT）启发。
Unified DeepSeek Harness plugin: role-based subagent routing + per-agent evolution — prefercmd/memory as knowledge allow/deny lists, so repeated tasks start from proven commands and save tokens
Native DeepSeek API balance and local token usage dashboard for DeepSeek Harness
DeepSeek Harness 原生视觉 Bundle：粘贴或拖入图片，通过托管的 deepseek-vision-mcp 调用 OpenAI 兼容视觉模型。
Extra prompt slots (${provider} ${thinking} ${time} ...) for the official turtle-ui TUI of DeepSeek Harness
Read-only security & compliance toolkit for DeepSeek Harness: prompt-injection detection (rule engine with a pluggable model classifier), Chinese-PII redaction, and a local configuration security audit that emits redacted, reproducible risk reports.
RAG 检索增强生成：向量数据库、嵌入、语义搜索、减少幻觉、来源引用。受 wshobson/agents（38k★ MIT）启发。
Sidebar usage card for the dsh web GUI: 7d/30d token usage, spend and remaining DeepSeek balance, rendered above the settings button.
DeepSeek Harness plugin: LLM model selection & deployment analysis (deployability, VRAM, TTFT, latency, throughput, power for 38 models × 20 GPUs/NPUs)
DeepSeek Harness plugin for progressive, source-grounded deep research, writing, and learning: steerable checkpoints and citations verified against retrieved sources
Indirect prompt-injection guard for DeepSeek Harness: taints tool output by origin and gates privileged tool calls that follow untrusted content
DSH 插件：把会话轨迹变成可分享、可复盘、可教学的结构化报告（脱敏 + LLM 总结，HTML/MD/JSON 输出）
Multi-agent orchestration for DeepSeek Harness: delegate, adversarial review, diverse explore, background runs + converge. One prompt. A team of agents.
常驻视觉服务：直连视觉模型（默认 opencode-go/minimax-m3，回退 zai-coding-cn/glm-4.6v）。describe_image / subagent_vision 工具 + 粘贴图片自动转译（llm/stream 钩子）+ 输入框视觉状态小胶囊与详情页（活动日志：指令/思考过程/输出）。零子代理、零 agent 上下文开销，按会话记忆窗支持视觉追问与验收。
DSH client plugin: Think token-count display + optional tool-call grouping.
Native DeepSeek Harness model health monitor with persistent Alive/Dead state, runtime failure verification, scheduled rechecks, and health-sorted searchable model pickers.
让 dsh 用独立的 OpenAI 兼容视觉模型读图：主对话历史只保留纯文本描述，图片字节不进上下文，纯文本模型也能读图。
仓库知识图谱：把代码/PDF/Markdown/截图/白板照片提取概念关系建图——god nodes/惊人连接/建议问题、EXTRACTED/INFERRED/AMBIGUOUS 边标签、71.5x token 节省、wiki 导航、--watch 自动同步、git hook。受 safishamsi/graphify（48.7k★）启发。
Display-only privacy mask for DeepSeek Harness chat: regex → ***, hover to reveal; copy and LLM context unchanged.