플러그인

Keeps every model request inside a per-provider allow-list, enforced at the agent/request waterfall: disallowed models are redirected to a fallback or refused; subagents stop inheriting stale creation-time routes; warns when consecutive requests would make a --models-max 1 llama.cpp router reload models. A one-entry list is a global single-model mode.

Kimi Coding web search and web fetch providers for the DeepSeek Harness — one coding-plan key fills both ctx.web seams (POST /v1/search and /v1/fetch), a settings card in the GUI stores it via the credentials service, and the fallback chain reads the Kimi CLI config for zero-setup installs.

Adds an http_probe tool for endpoint health checks: status code, latency and key response headers, using HEAD with GET fallback; response bodies are never downloaded.

Shows MiniMax Token Plan usage beside the DSH Web UI model selector with compact 5h/7d bars, click-to-refresh, automatic idle and heartbeat refreshes, and endpoint fallback. Reuses subscription keys configured under Settings → Model and appears only for MiniMax Token Plan providers.

All-in-one inline media previewer for mip-img/mip-video/mip-audio fences with streaming playback, ffmpeg transcoding fallback, lightbox, and trust-fenced LAN/mobile access.

Drag or paste any local file onto the DSH page; it resolves the real absolute path (Spotlight + size/mtime + head/tail fingerprint match) and inserts it into the composer, with a content-cache fallback.

Auto-switch the DeepSeek route to the vision model on demand: flash main session switches (A), pro keeps deep reasoning and delegates image reading to a vision subagent (B), subagents always switch, with fatal-failure fallback. No manual model switching.

Tiered compaction hub: replaces the single 80% compaction cliff with a configurable ladder (default 30/50/70/90) whose tiers escalate from gentle to maximum consolidation laws, folds a deterministic keyword index into every checkpoint so facts accumulate across cycles, and summarizes through a model chain — primary plus up to 3 fallbacks, each with its own reasoning effort. Adds a Compaction module sub-tab and a Settings card; toggling off always restores stock compaction.

Turns attached files into versioned evidence: `search_documents` builds a private local index (SQLite FTS5 after a startup capability probe, dependency-free JS fallback otherwise) and returns compact evidence blocks carrying an exact coordinate — PDF page, PPTX slide, text/DOCX line range, or quoted XLSX `Sheet!Range` — which `read_document` expands only while the content version still matches. Contiguous CJK runs are indexed as overlapping bigrams and queried as phrases, so word order is preserved; uploaded raster images take the native vision attachment path instead.

Routes dsh requests by conversation scenario and retries with fallback models when requests fail.

Subtitle-first video transcription and deterministic extractive Markdown summaries, with a faster-whisper large-v3 fallback when subtitles are unavailable.

MiniMax CLI bridge for DeepSeek Harness (dsh): mmx-backed web search fallback and transparent image understanding, with first-run onboarding and a settings card.

Plugin Cordis para o DeepSeek Harness v0.1: interceta a camada HTTP (auth barrier sobre o fallback da Web UI) e orquestra subprocessos de longa duracao com ciclo de vida reversivel.

DSH Tool Router — deterministically shrink model-visible tool schemas per step with a safe fallback

DeepSeek Harness OCR plugin — offline image text recognition powered by PaddleOCR-json (primary) and RapidOCR-json (fallback). 让 DeepSeek Harness 直接识别图片中的文字，无需视觉模型、完全本地离线运行。

Codex-inspired managed approval for DeepSeek Harness: risk-based MCP review, one-time grants, explicit denials, and human fallback on reviewer failure.

peon-ping sound notifications for DeepSeek Harness — a faithful port of the pi plugin pi-peon-ping-win (native Windows audio fallback, WinForms popup, OpenPeon sound packs). Beeps only when a task completes or terminates unexpectedly, by default. Settings

Two-tier image reading for text-only models: fast local OCR (RapidOCR, offline) first, vision-model fallback (modlens).

Automatic provider failover on rate limits or quota exhaustion: switch between providers with configurable per-provider models, a cooldown, and a permanent last-entry fallback.