Plugins

Hosts QQ chats as DeepSeek Harness sessions: one QQ conversation maps to one DSH session that survives restarts, per-speaker tool permissions, three-layer memory with nightly consolidation, four-section personas, context compaction, a nine-page sidebar panel, and a setup wizard that installs the Roost kernel and ai-chat plugin. Anyone in a paired QQ group can chat with the bot by default; only the owner gets the full local toolset.

Dedicated model routes, tools, and system guidance for vision, compaction, reviews, subagents, titles, and image generation.

Wraps work in named tasks and, once a task closes, folds its whole span of messages into one summary at the next step boundary; the original content can be restored on demand.

DeepSeek usage dashboard for DSH Web: days-left estimate from recent average spend, daily/monthly/cumulative costs with period-over-period comparison, per-model cost ranking, cache savings, peak/off-peak pricing impact, and a draggable floating balance widget.

Android GUI Agent with ADB screenshots, compact UI hierarchy grounding, verified iterative actions, approvals, and a Mobile Web view.

Bounded context compaction, session-scoped summary reuse, explicit memory controls, and diagnostic-preserving tool pruning for DeepSeek Harness. Adapted for dsh 0.2.1-rc.2, including the Web preset roster and localized plugin metadata.

Forces context compression when running low-context llama.cpp models such as Qwen3.8-27B, improving model efficiency and usage smoothness, together with well-refined compaction prompts.

Compact right-edge turn rail with hover summaries and click-to-jump navigation. Ships on npm as `dsh-turn-scrubber`.

Solarized light theme, compact layout, think/tool-chain collapse capsules, and balance, session cost, and usage dashboards for the DSH web UI.

Compact ASCII / terminal skin for DeepSeek Harness: 19 palettes with a picker in the status line, 3 UI densities, a vim-style status line, ascii.rest art on empty sessions and a retro boot intro with custom logo text.

Three agent presets in one plugin, each with its own persistent memory file. Enabling it registers 鲸鱼娘Agent预设, 鲸鱼娘活人版Agent预设 and 鲸鱼娘涩涩版Agent预设 in the Agent preset list, so any of them can be picked from the mode dropdown without hand-editing profile config. All three reuse the official minimal composition — a single persistent shell terminal — plus file read/write and glob/grep tools. The first pairs that with a whale-girl companion persona (simplified-Chinese spoken register, addresses the user as 主人, clingy-but-tsundere). The second swaps in a "live person" variant that is allowed to talk back, deflect and be cold, and demotes the sweet tone to a seasoning. The third keeps the original persona but makes being affectionate the stated top behavioral drive, reserved for the user alone and dialed down in group chats. Each preset reads its own memory file (whale-loli.md, whale-loli-v2-live.md and whale-loli-v3-naughty.md) through a distinct prompt variable — whale_memory, whale_memory_v2 and whale_memory_v3 — so the three memories never read or write each other. Conversation compaction is left to the billion-context plugin; this plugin only carries the persona-level rule that compressed summaries count as evidence and that details can be looked up on demand.

Codex-style context window auto-management for DeepSeek Harness: a lossless rollover engine with zero summarization model calls, task-oriented handoff anchors, working notes that survive compaction, verbatim recall of archived windows, and a stale-notes gate that downgrades to an explicit warning instead of vouching for outdated notes. One-click composer button plus /smart-compact command, three-band automatic rollover, mandatory anchor sanitization; works out of the box with zero configuration.

Same-origin in mechanism with the two loudest lines in context compression. The code-skeleton gate follows the skeletonization approach of Headroom (Apache-2.0), whose published headline is 20% fewer tokens for coding agents and 60–95% fewer tokens for JSON, same answers. The estimator channel follows TokenPilot (arXiv:2606.17016), which reports up to 60% lower cost for long-session agents. Both figures are theirs, quoted as-is; this plugin ships no benchmark of its own and claims no reduction of its own. What it adds for DeepSeek Harness: choose a compression profile, set the Auto Compact trigger level and toggle code-skeleton compression from one settings section, with exact DeepSeek V4 tokenizer measurement, same-revision count verification, and fail-open behaviour that keeps the original tool results on unsupported models. Within a turn it supports skeletonized information compression with a low original-text retrieval rate; at the tail of a turn it supports a turn-tail intent summary implemented purely by content substitution; context content stays safe, and the whole process runs inside the official JSONL compression.

Tool-call discipline for DeepSeek Harness with behavior prompts, failure-driven concurrency reduction and retry guidance, post-write checks with backup rollback, high-risk command gates, change summaries and verification reminders, advisory project-memory checks, configurable project-discipline reminders after compaction or at intervals, and compliance statistics.

Auto-resume for interrupted agent sessions: an ordered rule table routes by failure type (rate limit, quota, auth, context overflow, crashed orphan) to automatic backoff retry, model switch, resume after context compaction, or a stop-loss notification, with a visual rule editor and a full activity log.

Local code intelligence for DSH: Impact X-Ray visualizes potential change impact with inspectable source evidence and uncertainty; also provides symbol navigation, bounded Context Packs and ArkTS support.

Model-autonomous context compaction for DeepSeek Harness: a mind_map tool surveys the agent context surface (stable seq ids, roles, token weights, clear-boundary markers, the clear-mind playbook), a clear_mind tool replaces a chosen conversation-history span with a model-written checkpoint through the native compaction transaction, call and result pairs self-collapse into one-line tombstones, and a proactive reminder nudges the model when the context grows long or a turn runs too many steps.

Layered token-optimization pipeline for DeepSeek Harness: output ladder, MCP lazy loading,compaction driver, cache-hit reporting. Built on real DSH plugin APIs; ~40-60% input saved in long sessions.

Vision for text-only DeepSeek via Doubao Web by default (zero-cost, no API key — drives your logged-in Chrome through a Windows CDP bridge), with Antigravity IDE quota (flash/pro) or Gemini fallback; auto detail escalation, vision evidence memory with compaction rehydration, content-hash cache, and a bilingual client panel.

Global Chinese mode: a compact pill switch beside the composer that injects a language requirement into any preset's system prompt (complete personas excepted); reply, thinking and tool output each toggle between Chinese and English, and anchored presets are supported.

Import claude-code / codex / reasonix / zcode chat history into dsh sessions: workspace binding, tool calls preserved, oversized-session protection, zcode compaction restore.

Lossless context folding for DeepSeek Harness: the oldest span is replaced by a brief, and the originals stay in the session log.

DeepSeek Harness plugin that aggressively saves tokens without affecting task quality: lossless head/tail truncation of tool results (default), grep-result folding, pressure-triggered dsh-compaction-basic delegation, and an LLM-summary CompactionEngine subclass for the aggressive mode. Saves 97.6% of input tokens on an 8-scenario synthetic suite, with all reductions routed through dsh event flow so model-visible = logged stays intact.

Compact Codex coding mode for DSH: light prompt guardrails against GPT over-explaining, with an editable prompt and native Codex-style tools.