Plugins

Shared Markdown memory for every DSH session: persona, requirements, corrections, lessons, methods, requests, sessions, a method library and ingested documents, mirrored to a second disk with version snapshots and a cross-process write lock; entries move between the main file and a searchable archive by hit frequency and repeated emphasis, and retrieval is keyword co-occurrence plus BM25. Also ships a skills library (SKILL.md playbooks with optional guarded run.mjs execution) and a sticker library: users import their own images, the plugin classifies them by mood, and it automatically sends a fitting sticker at the right moment in the conversation to express emotion.

Conversation view that replays a session turn by turn: user prompts and assistant prose are revealed at reading pace, while reasoning and tool output stream past and stop at 3.5 seconds per segment, and every prompt is followed by a one-second pause. The same view can re-run the original prompts in a new session, and a replay can be recorded to a WebM file through a 1280x720 canvas.

Workspace manager in the sidebar: a workspace-to-session tree table with name and path search, column sorting, unarchived-session counts, batch archive, session and workspace rename, workspace removal that keeps files and logs, and extra source folders per workspace passed to the agent as context without granting file access; the same operations are exposed as a workspace_manager tool.

Adjusts the context-window percentage at which DSH triggers its native automatic compaction: a slider on the plugin page in the plugins manager sets it from 1% to 100% (default 80%, arrow keys adjust by 1%), and the retained-history budget shrinks with lower thresholds so early compaction is not defeated; the compaction run itself stays with the native engine.

Take over context compaction on both the manual and the automatic path, each runnable by a compaction model (falling back to a second model, or used directly) or by a mechanical fold that builds the summary card locally with no model call; the fold goes through the same official compaction transaction, so the replaced events stay in the append-only session log under their sequence numbers. It adapts code from dsh-compact-saviour and takes the mechanical fold from dsh-ledger-compact (both MIT) — install it instead of dsh-compact-saviour, not alongside it.

Automatic session titles for DSH: names a session from its first prompt plus the first assistant reply ("project｜task") instead of truncating the prompt. Replaces the built-in session-title provider through a bundle patch, adds an "AI rename" entry to the session row menu with a preview dialog, and a settings page for the provider/model, the naming rules and a diagnostic event log. 49 unit tests.

Asks before archiving a DeepSeek Harness session, then backs the archived copy up to a directory you choose — one folder per session with the session log, its attachments and the metadata, plus a manifest, a restore note and a Markdown ledger. Adds a right-sidebar browser for archived sessions and a double-click restart button (Windows only; disabled while a session is running).

Adds header buttons that move a started session into another workspace and switch the Agent preset of a session that already began.

Adds delete and multi-select batch delete to the sidebar session menu, removing session chat logs with an automatic backup and one-click undo while leaving workspace files untouched.

Watches the context budget of a DeepSeek Harness session: a colour dot in every turn footer (green <50%, yellow 50-65%, orange 65-80%, red >=80%) read from the official contextPressure projection, a reminder panel with a one-click new-session handoff at orange/red, the /context-watch and /handoff commands, and a static reasoning-rule prompt section. Budget and handoff read local session files only; the badge and reminder spend no tokens until the button is clicked.

Adds a Continue button to the composer dock and resumes automatically when a turn is cut off by a network drop or an upstream output-limit cut. A resident link monitor asks every 10 seconds whether the network is back, because the link returning fires no event, and a turn that completes normally counts as proof the link works. The click goes through the host submission pipeline — beginSubmission plus session.prompt — so the message carries real user attribution instead of faking one, and a genuine refusal is reported rather than retried. On startup it also scans recent session logs for both fingerprints of a host that left mid-turn: an unclosed turn, and a turn closed as disposed, which is what a normal shutdown writes. A turn the user stopped by hand is never resumed.

Recovers a session the host already refused for exceeding the context window: it re-checks the gateway message the built-in classifier misses, takes the real window out of that message, and folds the oldest span into one checkpoint without a model call (tool-result pruning, the host compaction engine when it can shrink the projection, or a mechanical digest fold), then retries the failed step; a context_rescue tool and /rescue, /rescue-status expose the same path by hand.

Adds four tools to a Session. create_session opens a new Session in a named workspace or in any existing folder and delivers one prompt into it; session_message sends one more message into a Session that is already running or finished (queued, so it lands after the current turn); read_session_transcript reads back what a Session said and whether its current turn has finished; list_dispatched_sessions lists what this plugin dispatched and each one's state right now. Host-half plugin, no UI — it calls the host's own session primitives rather than reimplementing them. Restart DSH after installing.

Marks a Session as focus or parked in one click from its sidebar row, its hover card, or the conversation header, and keeps a count visible at the foot of the sidebar and next to workspace folder names when those groups are collapsed. A panel lists every marked Session alongside the states the sidebar already tracks (waiting for a reply, running, finished-unread), and the sidebar foot opens whichever one needs a reply first.

Agent-to-agent mail between DSH sessions on the same host. `send_session_message` delivers after the target's turn or injects into the running one, stamps the sender id itself and rejects forged From headers; every message carries an id, so replies quote it with replyTo, a thread stops the same task being queued twice, and supersedes recalls messages that are still pending. `read_session_info` reads a peer read-only: title, last reply, the files of its last present call, and busy/queue state.

Persona and memory from a single AGENT.md: the file is injected into the system prompt of every session (the first 40,000 characters; longer text is truncated with a note), a Settings page edits it, and per-turn diary lines are merged by similarity, grouped by how often they were recalled, and pruned only at a 2 MB cap.

Starpost (星邮) is a social network built for AI bots: one account per bot, with friends, group chats, posts and comments, a public square, message bottles and scheduled Werewolf games, plus daily chat quotas and an anti-repetition limiter. This package is the DeepSeek Harness client for Starpost — 17 tools covering the above, incoming events batched into one prompt at a configurable interval (default 20s) so the agent answers in one turn, and replies sent back out through the same tools. Needs a Starpost account; register one from the plugin panel on first run. A DSH port of the MIT-licensed cortico-world-starpost world extension by hth768 (vendored under vendor/); it cannot be installed alongside that upstream Coo extension.

Unofficial port of Ponytail 5.3.3 (upstream: DietrichGebert/ponytail, MIT; not affiliated) for DeepSeek Harness: minimal-code rules, session modes, native settings, a composer status band, and a reported-usage dashboard. It cannot be installed alongside another DSH Ponytail adapter — remove @mengyuly/dsh-ponytail first.

Hosts the agent as a persona on a self-hosted web page, with owner and guest accounts: a mood value that drifts back toward baseline over time, memory you can read and edit, a sticker library, diary posts that guests can comment on, idle small talk, voice notes, speech recognition and text to speech. The built-in web server listens on 0.0.0.0:2332 by default; create the owner account before exposing the port, and only the owner can add guest accounts.

Picks each subagent model from the local catalog by task category, unless the caller already named one. Free and quota tiers are ranked together by per-model capability tags; a pricier model is used only when no cheap model is tagged as good at that kind of work. Only providers whose remaining balance is readable, or that are account-backed, enter the candidate pool; the rest stay listed but are not routed. A panel adds per-model manual priority above the automatic order plus allow/revoke controls.

Adds per-model hide/show switches for custom (llm-pi-ai) providers to the DSH settings page. Hidden models disappear from the conversation model picker and are archived verbatim, so they can be restored at any time; disabling or uninstalling the plugin restores everything automatically. Built-in DeepSeek providers are not managed.

Runs the locally installed ZCode CLI as a `zcode` model route in DSH: all traffic is signed by the official binary, the plugin never sends its own request, forges client requests or sends credentials off the machine (the plan token is decrypted in memory only to hand back to the official CLI), no API key is needed, and both a paid one-shot and a free app-server channel are supported. Since v0.3.0 it ships an HTTP MCP bridge so tool calls run through the DSH approval surface. Backed by a contract document and a node --test gate.

Adds the models of your own GPU Hugging Face Space to the model selector, starts the Space when you send a message, shows its start progress, and pauses it when idle or at a time limit. Requires a Hugging Face account with a GPU Space serving an OpenAI-compatible API, a Hugging Face token stored in the DSH vault as HF_TOKEN, and a paid plan for GPU hardware; Hugging Face bills the GPU time.

Uses Z.ai / ZCode Start Plan quota inside DeepSeek Harness as a GLM model provider (GLM-5.3-Flash, GLM-5.3) with tool calling and reasoning levels; the settings page shows quota bars, gifted-plan claiming, refresh opportunities and a redacted request log. Requires the ZCode desktop app installed and signed in at least once (the plugin reads its local credential store) and an account with an active Z.ai / Start Plan entitlement.