Plugins

Real-time usage, cost and balance meter for DeepSeek Harness: official-API balance, cost accuracy within ¥0.02, per-turn / per-session / per-model cost ledger, live color-tiered token speed and a docked one-click popup.

Three-tier agent memory harness: L1 always-on CLAUDE.md, L2 index, L3 detail skeleton with templates, a gate that blocks writes to protected files without a one-time approval token, zero-judgement event logging, weekly distillation, and a TDD execution-loop skill.

Zero-touch runtime telemetry for DSH: every session auto-writes task, model tier, tools, failures, duration, status into a human-readable journal with a stats section (pure observer — no tools, no LLM calls, no injection).

The model switch for DSH: point it at any OpenAI-compatible endpoint, with curated free/cheap DeepSeek provider presets and automatic fallback when a free tier rate-limits you.

Six-layer local memory for dsh: captures turns, stages them through an L2 buffer into DL3 summaries via a staging model, injects a stable memory block each turn and replaces chat history with placeholders. Ships five memory tools and a config page.

OpenCode Zen free-tier models for DeepSeek Harness with anonymous no-key access, a loopback OpenAI-compatible shim, and real bash/read gate tools.

Model channel auto-enricher for DeepSeek Harness: eliminates manually configuring contextWindow, maxTokens, \[text, image] multimodal input, and {low, high, max} reasoning effort tiers model by model after adding a provider channel, with live auto-enrichment on save, a Settings UI page, and timestamped backups.

Trip-planning agent preset for mainland China that answers in Chinese: queries 12306 train availability, Fliggy and Tuniu flight and hotel prices, and real Amap transit routes, checks closures and weather, and delivers a single offline HTML roadbook with draggable maps, three hotel price tiers and an itemized budget. Ships 9 skills, an environment doctor that walks users through missing API keys, and a patch for the 12306 MCP server; it plans and queries but never books or pays.

Replaces the reasoning-effort dropdown with a draggable snapping slider, and draws 13 canvas effects across five style families (thrust, fluid, fantasy, tech, texture) with three tiers that unlock as the effort level rises. Custom mode combines 10 particle forms with 8 flow trajectories; 9 accent presets and a hex picker recolor every effect, and optional key sounds cover two voice clips, a synthesized duck squeak, a mechanical click and a local audio file.

A native reasoning-effort control for the composer: a blue-to-violet tier bar with a starfield and a sheen sweep. It owns the model seat and drives the official model-directories service, so the ladder comes from each model's own reasoning declaration and no other plugin is required.

SearXNG search and Scrapling multi-tier fetch provider bundle for DeepSeek Harness ctx.web with SSRF protection and markdown conversion.

Windows system-drive (C:) cleanup for DSH: rule-based scan, five-tier safety grading, a staging area with a rollback ledger, UAC elevation, and migration to another drive — a model tool plus a Web GUI panel.

Image and video production toolkit for DeepSeek Harness - background removal with quality tiers, local inpainting retouch, headline and logo compositing, batch platform-size exports and image QA reports; for video, AI subject matting with background replacement (RVM ONNX, plus a chroma-key path), animated titles with fade/slide/typewriter/karaoke presets, 53 transitions with stylized variants, and light-effect and grading presets with LUT support, alongside footage inventory, accurate cutting, subtitle burn-in, music mixing and platform-spec exports; it provisions its own local Python runtime through a bundled bootstrap script.

Unified model routing for DeepSeek Harness: one logical ModelID over multiple providers with first-token failover and cooldown, health-aware candidate ranking, three tiers (tier1/tier2/tier3) auto-selected by purpose, per-candidate reasoning effort, and a settings-page management panel.

Reasoning-effort tier manager for custom API models plus cross-session preference memory: inline low/medium/high/max toggles inside Settings → Models, with atomic settings writes.

A consent-gated personal profile store: people are indexed by name, entries by person and key, and no profile record is read before the current session is authorised. Sensitive entries come back masked unless the user authorises that tier separately; credential-like values are refused at write time. Ages and next birthdays are computed at read time, so asOf restores the answer for a past date, and facts that rot (height, weight, address) carry freshness warnings. Five profile_* tools plus a settings page to browse, edit, delete and revoke grants.

Long-term memory for dsh with defense-in-depth hardening. Two tiers (project `.dsh-memory/` and user `$DSH_HOME/memory/`), each a plain hand-editable JSON file with bounded entries and tags. Writes are atomic (unique temp file + rename, previous revision kept in a bounded history/) and serialized across processes by a per-tier lock with stale reclaim; a disk-revision check refuses lost updates instead of silently overwriting another writer. Memory text is wrapped in explicit untrusted-data boundaries (forged wrappers stripped on write) to blunt prompt injection from remembered content, load-time validation rejects out-of-bounds stores, and symlink escapes out of the tier directory are refused with MEMORY_STORE_ESCAPE. No content-heuristic filtering and no confirmation step, so headless auto-consolidation stays intact. Install: dsh plugin --profile web add github:RocoFish/deepseek-memory.

Four-tier prompt enhancement for the dsh web GUI: base / standard / professional / system — the system tier always emits a LangGPT-structured Markdown prompt, the iterate workbench rewrites an editable base under an optional requirement, and literal ancho

Kifu-style context compression for long DSH sessions: each step folds new turns into a DAG (node = position, edge = attempted fix), injects only a ~415-token hot list (2000-edge fixture, L5 tier), and adds failure memory plus position-repetition detection.

RunningHub cloud ComfyUI workflows as a background task pool: submit a batch, and a watcher dispatches it within a concurrency budget, polls each job, downloads every output to disk, and wakes the session when the pool drains. A shared ledger records batches, costs, instance tiers, submit and finish times, and per-file save paths, with a live board in the right sidebar.

Bilibili curation floating panel for DSH: learns your interests on a probability schedule and decays them by half-life, picks videos inside a token budget derived from a daily money cap, and mentions you only when a fit score passes a threshold. Comments are off by default; once enabled they run in draft-and-confirm or self-checked auto mode under a daily quota, with a hard no-comment rule for political, government and frontier-science topics, and comment text is scrubbed for account identifiers before it can be posted. It signs in through your own Bilibili account cookies; a separate posting account is recommended.

Connect a self-hosted Magpie gateway with one provider route per upstream owner, so the stock model picker shows Magpie (Claude), Magpie (Codex) and other groups, plus automatic multi-line failover, read-only usage cards, per-model default effort, a per-session fast switch for priority-tier models, and optional Cloudflare Access.

Native generation-time Mixture-of-Agents: every turn, reference models run the same session context in parallel as pre-run drafts and one aggregator executes with full tooling; config-as-model recipes, F mirror / R trimmed views, enforced subagent gating (dual-shape session indexing), derived per-recipe vision (vision refs relay image reads to text-only aggregators), tiered cadence with separate steering and patrol rosters plus frozen reuse between beats, degeneration guard dropping degenerate reference output, anonymized model face (failure notices carry no provider identity).

Adds a tier bar to the DSH composer with three thinking tiers (normal, strict, max), an exclusive loop mode whose three levels run 3, 5 or 8 self-review rounds, and a counter-review toggle that stacks on either; the model reads the current tier through the clarify_tier and forge_level tools.