dsh-neocortex
IdentifiedZhinao (second brain) for DeepSeek Harness: Markdown-authoritative layered memory (L1 long-term model gated by user confirmation / L2 knowledge with mandatory sources / L3 recent status under daily qu
dsh-neocortex
Zhinao (Neocortex) — a Markdown-authoritative second brain for DeepSeek Harness, with a local recall ledger, lifecycle engine, weekly review and a Web panel.
Features
- Layered memory (L1/L2/L3): L1 long-term model (background, collaboration preferences, task standards) gated by user confirmation; L2 knowledge written automatically with mandatory sources; L3 recent status written directly under daily quotas.
- Local ledger with recall explanations: a crash-conscious local ledger accounts for every recall and write; the Web panel explains why a memory was injected.
- Lifecycle & weekly review: scenes and entries sleep, wake, revive and get promoted by bounded rules; a resumable weekly review queues L1 candidates and sleeping entries for the user.
- Web panel: state, recall explanations, L1-only confirmation, review actions, read-only L2 audit, undo, and six bounded settings sliders.
- Blank template initialization / takeover: first run initializes your brain from a generic, personal-info-free Markdown skeleton, or adopts an existing Zhinao directory without touching its files.
Requirements
- DeepSeek Harness (DSH)
>= 0.2.1-rc.5. - Verified on Windows.
Install
dsh plugin --profile web add dsh-neocortex
or install dsh-neocortex from a DSH plugin market in the GUI.
First run
Open Settings → Plugins → Zhinao / Neocortex and run the setup wizard. You choose the directory yourself in every mode:
- New brain: pick an empty directory; the wizard copies the blank template into it. Any conflicting file aborts the initialization — nothing is overwritten.
- Use existing: point to an existing Zhinao directory; the wizard only verifies it and never modifies existing files.
- Upgrade skeleton: point to an existing but incomplete Zhinao directory; the wizard creates only the missing skeleton files and skips every file that already exists.
Usage
Agents get three tools:
neocortex_remember— write one atomic fact to L1/L2/L3 with quota or queue feedback.neocortex_recall— recall general and sleeping personal history; routes external-knowledge questions to web search.neocortex_panel— panel state, review preview, undo, review execution/retry, and bounded settings updates. L1 confirmation and weekly-review accept/reject stay user-only.
You can also steer it in natural language: "skip the brain this time" (no read/write), "don't remember this" (don't save), "remember this" (save it).
Uninstall
Removing the plugin does not delete your brain data: the Markdown files, the local ledger and settings stay on disk.
Development
node --test tests/*.test.mjs
For learning purposes.
Comments
Loading…
Similar plugins
by jipika
Two-layer long-term memory for DeepSeek Harness (DSH): global + per-project markdown, live re-read, zero extra LLM cost, with a Settings panel. 给 DSH 的两层长期记忆。
★ 6
↓ 656/wk
MIT
JavaScript
Oct 27, 2026
dsh plugin --profile web add @jipika/dsh-memoryby Frog755
Three-layer local memory for DeepSeek Harness: L1 snapshot memory (MEMORY.md/USER.md) read on demand through the memory_* tools + L2 searchable knowledge base (facts + SQLite FTS5, Chinese search via
★ 3
MIT
JavaScript
Sep 10, 2026
dsh plugin --profile web add @frog755/dsh-hybrid-memoryNeuron-inspired memory for dsh: turns buffer into a hippocampus, a nightly patrol classifies fragments into pools, and each turn injects a minimal auto-compacted handoff, the live mood (via dsh-feiyu-
★ 0
dsh plugin --profile web add dsh-neuron-memoryUser-authored tiered long-term memory for DeepSeek Harness: injected as a system-prompt section, with a Chinese panel to view and edit it. The AI never writes to memory on its own.
★ 1
MIT
JavaScript
dsh plugin --profile web add dsh-user-memoryby menotbobbybrown
Persistent long-term memory for DeepSeek Harness: a JSON store of typed entries (fact, preference, entity, rule, episodic) exposed as the memory_remember and memory_recall agent tools, with recall ran
★ 0
MIT
TypeScript
Sep 8, 2026
dsh plugin --profile web add @modelnorth/dsh-plugin-memoryby hanhan9999999999999999999
Local-first memory nebula for DeepSeek Harness: graph, recall, paths, hygiene, snapshots.
★ 5
MIT
JavaScript
Nov 18, 2026
dsh plugin --profile web add dsh-memory-nebula