Plugins

Git-native, auditable long-term memory for AI agents

Probes Windows-host Ollama, LM Studio, vLLM, and llama-server from WSL, compares ctx, and reports /v1/models apiReady versus TCP.

Reports Docker CLI, context, and daemon reachability in WSL, plus vLLM :8000 /v1/models health and GPU runtime hints.

Web search provider for the ctx.web seam that routes the DeepSeek web_search tool through a LiteLLM proxy using the OpenAI Responses API, reusing LITELLM_API_KEY.

Memory consolidation tuning for dsh-mneme (autoDream enabled, schema-max output budget, deepseek-chat route) plus the llm-deepseek reasoning-effort fix, an MCP bridge to the auditable seed-society agent runtime, and six yogacara seed skills.

Governs multi-model access, monthly credits, manual/quality/credit/auto routing, failure rerouting, and usage auditing for DSH.

Command Code LLM provider plugin adapted for DSH-Desktop 0.7.1: registers a commandcode route, with a live model catalog, reasoning-effort support, and real-time plan usage display.

Three independent language channels for model output — tool-call descriptions (every tool-call card title, including run_code in PTC-style presets), thinking, and replies — each set to follow the web GUI language, force a specific language, or turn off (default: only descriptions, following the GUI). One global runtime-context prompt directive on the host; the client half reports the GUI language; a Settings plugin card switches the modes.

OpenAI-completions-compatible adapter for custom gateways (vLLM / LM Studio / self-hosted proxies): always role:"system", thinking driven by the model config, Qwen-style response split, and vision-model image input (single / multiple).

Plugin-button aggregator with carrier + on-site-LLM extensibility: gather any plugin's scattered buttons into one entry (SSiD title bar / web floating ball), LLM registers per environment, right-click deregisters — pure data, no plugin source touched.

Per-provider concurrency gate on the llm/stream waterfall: holds surplus model requests in a FIFO queue inside DSH so main agent, subagents and compaction take turns on a single-slot local server (llama.cpp --parallel 1) instead of timing out; maxConcurrent / maxQueued / queueTimeoutMs.

Ask a sidebar assistant about selected assistant text: an ephemeral right-side panel answers from a snapshot of the main conversation's LLM history with streaming Markdown, collapsible thinking and multi-turn follow-ups, without touching the main conversation.

Stavros RedTeam for DeepSeek Harness (DSH) — authorized red-team persona + scope-guarded pentest tooling (persona, 24 subagents, methodology and 78 zero-dependency guard tools).

Routes the harness through the OpenLLM gateway: a pure-config Cordis patch that adds an `openllm` provider to the in-box pi-ai adapter pointed at the local daemon (`127.0.0.1:8787/v1`, no API key held by dsh) and registers the `openllm mcp` stdio server (openllm, claude-context, supermemory tool groups).

Provider/model-aware adaptive pacing and visible rate-limit retries for DeepSeek Harness free models.

Fail-closed network recovery, recommended answers, and safe unattended continuation for DeepSeek Harness.

Turns code review into a hard gate: deterministic severity rules, LLM-assisted findings, a team approval quorum and a durable compliance audit trail that must pass before a merge proceeds.

Piano performance plugin: ask the agent to play a piece and it renders on a Canvas2D grand piano with real Salamander Grand samples, an immersive stage, and an interactive 88-key keyboard.

让 DeepSeek Harness 自动重试商汤 429/QUOTA 等瞬时网关故障直到任务完成——启动时给所有 llm-pi-ai provider 热注入加固版 retryPolicy。

Uses an isolated LLM review to approve or reject DSH sandbox permission requests.

A pouch of practical micro-plugins for DSH: per-turn wall-clock time injection, safe restart, session ID header badge, Tavily search provider, configurable LLM request-header routes, and an MCP server console managing the official @deepseek-ai/dsh-mcp-client at runtime — add/edit/enable/disable/reconnect, per-tool switches, live status over SSE, mcpServers JSON import, and auto-migration of profile-YAML MCP entries.

Run the main session model on a local Claude Code subscription through the official Claude Agent SDK, taking over the llm/stream route for the claude-code provider, with token-level streaming, session resume, and DSH tools bridged in over MCP so they keep DSH sandbox and approval.

Multi-language subtitle translation workflow for dsh: SRT/VTT parsing, sentence-level LLM translation, bilingual merge, batched with resumable checkpoints.

LLM approval answerer for sandbox escalations beyond workspace-write: a deterministic regex filter first, then a clean-context reviewer model; humans are asked only when it is unsure. Requires the bundled core patches.