プラグイン

Amazon product research workbench for DeepSeek Harness: category, brand and follow-seller collectors driven through the local Edge browser, a product library with unified filters, an FBA/FBM/self-operated profit calculator, blue-ocean category scoring with candidate picks, trademark lookup, supplier image search and Excel export. Proprietary software: it installs and starts without a license, but every data call returns 403 until the author issues a device-bound license key. Needs Edge started with --remote-debugging-port=9222 and signed in to Amazon; starts a local backend on port 3088. The separate 智赢 collector browser extension is optional and adds size, weight, trademark and monthly-sales fields.

Memory engine for a long-lived companion agent: a six-module loop (gate, solidify, store, retrieve, inject, decay) that decides what is worth keeping, distils raw turns into long-term entries, fuses short and long term at recall time, and fades entries that stop being used. Runs offline by default; if DEEPSEEK_API_KEY is set it also sends each message to the DeepSeek API to judge and distil it. Zero runtime dependencies and no build step; ships three switchable panels over the same store - neural brain, memory ring and memory galaxy - so you can watch what it remembers. Non-commercial license: commercial use needs the author's permission.

Permanently deletes one session from its sidebar context menu after a confirmation dialog: the log directory under the DSH home, the projection-cache thumbnail, its id in every workspace.json, and its archived and pinned entries. Sessions resumed since launch are detached from both the sessions and the agents stores first, so the deleted row does not return to the sidebar as an orphan. Refuses while the session is running a turn, and never touches the workspace folder itself.

Continues every session whose turn was left open by an unexpected DSH exit.

Adds a Delete conversation action to the conversation menu that permanently removes the session log, projection cache, and sidebar entry; refuses while a turn, subagent, job, or reminder is running.

A Conductor tab and commands for a chat's context: /new replaces the context in place with a model-written handoff, /rewind restores the chat and the agent's file edits to an earlier prompt, plus /drop, pins, an automatic compaction level, recovery after failed turns, silent messages between chats and agent and task orchestration. Supersedes dsh-orchestrator, dsh-force-compact and dsh-continue, so do not install it alongside them. Requires DSH >= 0.2.1-rc.5, Node >= 22.5, the web profile and dsh-better-sidebar.

Tiered cold storage for DSH sessions: idle sessions move out of the live root into an on-disk cold tier instead of being deleted, an FTS5 index over that tier keeps them searchable, and a restore path moves a session back into the live root. Dry-run by default: nothing moves until you turn it off in Settings.

Merges the work done in a forked session back into the session it came from: /merge turns the branch's session log into a digest without a model call (each turn's prompt and reply, files changed, commands run, choices you made, to-dos and the final answer) and delivers it with your next message, or at once with --wake; merging the same branch again brings only the turns added since. Also adds /branches, a right-sidebar branch map with merge state, a reminder above the composer for branches with unmerged turns, and session_branches / session_branch_merge tools for the agent.

Switches the tone of DSH replies from Settings → 人格模式 or the /persona command (/persona catgirl, /persona off), effective from the next turn. Only the active persona is injected, as runtime context, so the system prompt keeps a single stable short strategy; with the mode off the injected persona text is 0 characters. Facts, code and engineering discipline are unchanged.

Per-turn reasoning-effort routing that keeps the current provider/model: judges each user message (retry or correction, explicit intent, length, code blocks, errors, task scope) and sets reasoningEffort to a level the model actually declares, so Magpie and other OpenAI-compatible models work too. Ships a composer chip with an auto/manual switch and per-session decision history.

Replaces the composer model picker with a model + thinking-effort picker: arrow keys left/right cycle the effort level and arrow keys up/down switch model; an Auto level scores each turn by how far the conversation has drifted from the main objective and by repeated failures (one extra small judge call per turn), picks the effort from that score, and labels every finished turn with the effort it sent.

Sidebar token and cost curve for the current conversation plus a local ledger across all sessions: monthly/weekly rollups, per-turn detail, Excel export with print-ready A4 layout, and a live account balance. Reads session logs only, prices every message by its own peak/off-peak timestamp, and drops fork-inherited history so it is never billed twice. Desktop app only, verified on Windows; the live balance needs a signed-in DSH account.

Estimates recorded DeepSeek model spend per conversation and turn, with a cost-ranked overview. Built-in tariff table covers 2026-09-11 to 2026-12-31.

Bottom-right whale widget showing DeepSeek balance, today's spend from the official usage API, per-turn cost narrated after each turn, and a peak/off-peak pricing countdown, with six expressions, click-to-refresh, weather and a bilingual (中文/English) UI.

One loopback GET request returns the DSH account wallet balance (recharge + bonus), the same figure as Settings → Account.

Per-turn cost meter and tool-loop guard for the DSH web GUI. It records model, tool and idle time for every turn, blocks runaway tool calls with a dynamic step gate, a retry breaker and a slow-command block, and renders the breakdown as a card under each reply plus a ledger view.

Models 64-token cache blocks and calculates per-turn API costs and cache ROI.

Records token usage per turn into two ledgers, imported history before installation and new events after, and shows exact per-bucket totals, per-model and per-session breakdowns, and day/week/month filters in a sidebar panel.

Turns a live session into a layered instrument: seven toggleable rings (session, steps, skills, tools, jobs, approvals, file changes), a per-turn timeline, hover and click on every ring node, and per-step ladders drawn beneath the rings. It reads the session log directly and updates as the conversation runs, in Chinese and English.

3D office view for DeepSeek Harness: the sidebar panel turns the real main session, its sub-agents, tool activity and reports into animated desks, with per-desk screen details and a notification centre for collaboration events; sessions are picked per working directory. Read-only: it starts no tasks, calls no model, needs no API key, and ships its own scene and models (no CDN). Requires DSH 0.2.2-rc.1 (or a later 0.2.x) and Node ^22.19.0 || >=24.0.0.

A reading layer for long DeepSeek Harness sessions. A fixed enlarge button on every answer opens a two-pane view with 80-300% type, an outline of every turn in the session (turns not yet loaded are fetched on demand), and a composer that fills the real DSH input without sending. Formulas, tables and code stay exactly as rendered; over-wide formulas shrink to fit, failed ones degrade quietly, and turns whose answer was cut off are labelled as truncated.

Keyboard shortcuts for DSH Desktop: return to the conversation, focus the composer, switch pages, open DevTools, and an Alt+Tab style session switcher that previews sessions while Alt is held and switches on release (DSH NEXT on Windows).

A floating quick-ask panel that answers without touching the conversation: a 10-turn scratch window, question-and-answer favourites that do not take up a turn, the conclusion dropped into the composer on request, and an empty window on every restart.

Add an optional rejection reason to approval prompts: the text is steered into the running turn before the rejection is answered, and the panel only follows the session you are currently viewing.