dsh-clear-conversation

Say "clear everything above" and DeepSeek Harness actually forgets it.

Long sessions get slow and expensive: every request resends the whole history, and old topics leak into new answers. This DeepSeek Harness (dsh) plugin lets you drop that history in plain words, mid-conversation, without starting a new session.

What a session looks like (illustrative):

You:   My favourite colour is green. Now help me name a cat.
dsh:   How about "Basil"? It suits a green-loving owner.

You:   Clear everything above, then tell me a joke.
dsh:   [clear_conversation]  Context cleared.
       Why don't skeletons fight? They don't have the guts.

You:   What's my favourite colour?
dsh:   I don't know. You haven't told me in this conversation.

Quick start

dsh plugin --profile web add github:William20100221/dsh-clear-conversation
dsh web

That's it: no build step and no configuration. New to dsh, or want to try it safely first? See TESTING.md.

Features

• Ask naturally. "Clear everything above", "forget the previous chat", "start over", 清空上下文, in any language.
• Keeps your request. "Clear, then do X" still does X: only the turns before your message are dropped.
• Saves tokens immediately. The model stops receiving the old turns from its very next request, even within the same turn.
• Nothing is deleted. Old turns stay in the transcript and the session log; they just stop being sent.
• Hard to trigger by accident. Casual phrases like "forget it, let's move on" or "how do I clear the context?" are ignored.
• /clear command. Clear everything yourself while the agent is idle (Web and SDK).
• Safe by construction. The system prompt is never touched, and a clear that would split an unfinished tool call is refused.

How it works

dsh builds every request from the session's surface, an ordered list of messages. Its own /compact command replaces a run of old turns with an AI-written summary. This plugin uses the same mechanism, but writes one short note instead of a summary, so it costs nothing to run.

flowchart LR
  A[System prompt] --> B[Old turns] --> C["Your message: clear, then X"] --> D[Reply]
  B -. replaced by .-> N["Note: earlier conversation cleared"]

FAQ

Does it delete my messages? No. The transcript and session log keep everything. Only the model's view changes.

Can I undo a clear? Not yet. The old turns stay readable, but cannot be put back into the model's context.

Which dsh surfaces does it work in? The tool works everywhere the model runs. /clear appears where dsh shows commands: Web and SDK, not ACP or headless.

How do I remove it? dsh plugin --profile web remove dsh-clear-conversation, then restart dsh.

Status? Preview, for dsh 0.2.1-rc.5 and later 0.2.x releases. The clearing logic is covered by tests; a live run inside dsh is still being confirmed. Please open an issue if something breaks.

Other harnesses

This repo also keeps the earlier Cherry Studio work. Each part works on its own.

Cherry Studio chats: MCP server

mcp-server is a tiny tool you add in Cherry Studio's MCP settings. No fork needed, and it really saves tokens. It works by writing the Ctrl+K divider into Cherry Studio's database, with a backup first. If a Cherry Studio update changes the database, it stops and asks you to update.

• Needs Node.js 22.13 or newer. node server.mjs --check verifies your setup and tells you when a newer version exists.
• Chats only. Agents keep their memory in Claude's own transcript, which nothing outside the app can reset; use the fork for agents.

See the MCP server README for install steps.

Any skill-loading app: the skill

skill/clear-conversation is an Agent Skill. It works in stock Cherry Studio agents, Claude Code, and other apps that load skills.

• Install it from the latest release ZIP, or from GitHub with the link to its SKILL.md.
• Without special app support, it does a soft reset: the model ignores the earlier conversation and tells you the app's own way to clear for real.

See the skill README for install steps and phrases.

Cherry Studio fork

William20100221/cherry-studio, branch feature/clear-context-tool adds a built-in clear_conversation tool. With it, a clear is real: older messages stay on screen but are no longer sent to the model, which saves tokens.

	Chat topic	Agent session
Needs the skill	No, the tool's description is enough	Yes, for the full rules
Where the divider goes	Directly above the message that asked	After the current turn
When the fresh start begins	The next turn	The next turn, once the session is idle
Undo	Click the divider, or the Undo toast	The Undo toast (10 s)
Ctrl+K / eraser	Divider now, or a queued clear while streaming	Same
Ctrl+L	Stops the reply, then divider or delete (setting), queue continues	Same
Delete mode	Always asks first, also when the model requests it	Same

Settings → Context Management has two switches: the default clear action (divider or delete) and whether the assistant may clear at all.

In an Agent session, follow-ups already queued behind the clearing turn still run with the old context.

Development

npm test

The tests exercise the clearing logic against a stand-in session; they do not start dsh.

License

MIT for this repo. The Cherry Studio fork stays under Cherry Studio's AGPL-3.0 license.