Agent collaboration · DeepSeek Harness plugins — Page 260

Self-healing cross-provider model fallback for the DeepSeek Harness

Provider/model fallback chain plugin for DeepSeek Harness — when the current model fails with an eligible error code (e.g. RATE_LIMIT or QUOTA), the agent loop automatically retries with the next provider/model in the configured chain.

SQL 优化模式：EXPLAIN 分析、索引策略、N+1 解决、查询改写。受 wshobson/agents（38k★ MIT）启发。

DSH harness plugin: tolerate OpenAI-compatible providers whose streaming responses omit finish_reason (e.g. the Snowflake Cortex gateway) so turns complete instead of failing with a TRANSPORT error.

Kimi (Moonshot AI) chat-completions adapter for the DeepSeek Harness LLM seam — three routes: kimi-code (Kimi Code subscription), kimi-cn, kimi-global

LongCat (LongCat-2.0) adapter for the DeepSeek Harness LLM seam — 1M context, thinking mode, tool calling

DeepSeek Harness LLM User-Agent/header injector with key pool round-robin: keeps the official llm-pi-ai adapter, re-injects user-agent, and supports multi-key rotation per provider. | DeepSeek Harness LLM 请求头注入器 + Key 轮询：保留官方适配器，补回 User-Agent，支持按提供商多 Key Round-Robin 轮询。

DSH host plugin: retries EVERY model-gateway error (429, quota, auth, network, ...) with a linear 5s-step backoff capped at 5 minutes, covering agent-loop requests and direct ctx.llm.stream calls (compaction, session title). Max retries configurable in Settings; 0 disables this plugin only.

Volcengine Ark Agent Plan and Coding Plan providers for DeepSeek Harness

Generate llms.txt for DeepSeek Harness (dsh) plugin repositories from package.json + README — AI-readable discovery files with zero runtime dependencies. CLI + agent-callable llms_forge tool.

Local DBA tools for DeepSeek Harness: query, schema inspection, DDL/DML, backup/restore and slow-query analysis for MySQL/MariaDB and PostgreSQL

Local File Share / 本地文件共享 — let the dsh agent list, read and write files on the machine that runs your browser, through a private WebSocket relay and the File System Access API. Zero data copied onto the dsh host.

Host-owned long-horizon task status for DeepSeek Harness local coding agents (tools + inject + STATUS.md + rightbar).

Bounded-file memory for DeepSeek Harness local agents (ME.md + project MEMORY.md vault, cache-stable inject, inbox, Memory rightbar). No store, no embeddings, no LLM calls.

DSH plugin: discover this machine's HTTP proxy (proxy env vars, else the Windows system proxy) and make DSH use it — for the agent's in-process fetches and for every child process it spawns.

Watches that the UGUI制作模式 entry button materializes for ugui sessions and reloads the page once when it does not (cold-start boot-manifest gap).

Preset-scoped multi-Canvas uGUI Workspace with a standalone-only browser designer, extended selectable, scrolling, masking and layout components, stable node IDs, safe gesture undo, image staging, and explicit Unity Prefab generation.

Local web fetch provider: fetches web pages and summarizes/parses them with the local Ollama model before returning to the agent

DSH 本地通知栏：侧边栏【通知】入口 + 全屏通知中心（卡片列表/时间筛选/标题内容搜索/未读已读/删除/详情弹层一键复制），notify_add agent 工具与 dsh-localnotify CLI 写入，JSON 文件持久化，页面实时刷新。

Loom (织机) — second-validator plugin for DeepSeek Harness: a silent external coach that weaves your agent's tools, skills, config and even its model from usage history, with deterministic verification and cold-apply.

DeepSeek Harness plugin: a circuit breaker that denies the Nth identical tool call in a row, by exact hash of tool name and canonical arguments

Continue a DeepSeek Harness agent turn when the model narrated but did not call a tool.

死循环/重复输出检测插件 for DeepSeek Harness (dsh): 文本重复检测 + 工具调用意图级重复检测, 命中后自动 steer 重试 / cancel 强制中断

Result-aware stuck-loop guard for DeepSeek Harness: advisory nudges plus a monotonic hard stop. Only repeats with identical results count, so progressing polls and post-edit re-reads never trigger.