A MIT-licensed local Go memory tool for DeepSeek Harness and 19 other coding agents: BM25 over disk sessions, no LLM, no embeddings, starts full instead of empty.
DSH integration
Compatible
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-23
License
MIT
01What can it help you accomplish?
Give DeepSeek Harness (dsh) and other coding agents recall over sessions already written to disk, including months before install
A local memory index (BM25, no LLM, no embeddings) that returns the right past session when needed, with sub-millisecond lookups over GBs of history
DeepSeek Harness (dsh), Claude Code, Codex, Cursor and 17 other coding-agent users who want durable cross-session memory without re-debugging old fixes
02How to install into DeepSeek Harness
Not specified by the author
03DSH integration and capability boundaries
A local Go memory binary that indexes sessions coding agents already wrote to disk and recalls them across Claude Code, Codex, Cursor, DeepSeek Harness and 17 others
Disk-session memory index
sessions coding agents already wrote to local disk→the relevant past session handed back when needed; starts full, not empty
no LLM, no embeddings — one local Go binary indexes history already on disk
04Who is it for? When not to use it?
Good for
- DeepSeek Harness (dsh), Claude Code, Codex, Cursor and 17 other coding-agent users who want durable cross-session memory without re-debugging old fixes
05Compatibility, maintenance and safety notes
- Memory is built from sessions already written to this machine's disk; it is a local Go binary with no LLM or embeddings, and only recalls history the agents have already produced locally.
MIT · actively maintained (latest release v0.18.0, 2026-08-22)
06Frequently asked questions
How does deja-vu work with DeepSeek Harness?
deja indexes the sessions your coding agents already wrote to disk. When DeepSeek Harness (dsh) needs a past decision, the agent calls deja itself and gets the relevant session back — nobody has to search manually.
Does it need an LLM or embeddings?
No. deja is a single local Go binary using BM25 full-text search. There is no LLM and no embeddings, so it runs fully offline on your machine.
Does it start empty like other memory tools?
No. Every memory tool starts empty and records forward; deja starts full because it indexes the history already on disk, including sessions from months before you installed it.
How accurate and fast is the recall?
It reports 85.3% hit@1 on LongMemEval-S, 69.6% on LoCoMo, and sub-millisecond lookups over 5 GB of history.
Which agents does it support?
Claude Code, Codex, Cursor, DeepSeek Harness and 17 other coding agents that write sessions to disk on the same machine.
07Related DSH workflows
openviking
by volcengine
Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
colleague-skill
by titanwings
Turn memories and knowledge into warm, reusable agent skills — a skill generator for digital companions in the AI persona era.
nocobase
by nocobase
NocoBase is an open-source AI + no-code platform for building business systems fast. Instead of generating everything from scratch, AI works on top of production-proven infrastructure and a WYSIWYG no-code interface, so you get both speed and reliability.
archify
by tt-a1i
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.
08Data and sources
Memory for coding agents, starting with the history you already have.
deja indexes the sessions Claude Code, Codex, Cursor and every other agent on this machine already wrote to disk
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-23. Found an error? Submit a correction.