🧠 Mneme

A cue-indexed tag memory for conversational agents — no embeddings, no LLM in the memory path.

Most agent memories embed every turn and distill it with a write-time LLM, then search by vector similarity — expensive and opaque. Mneme instead does what human recall does: index by cues, reinstate by cue-match, and spread activation along both shared meaning and the thread of conversation — with a plain inverted index over deterministic tags. No vector database, no language model, no embeddings.

At zero memory cost, under an identical controlled comparison, Mneme matches a strong dense retriever on LoCoMo (F1 52.6 vs 52.9) and beats it on high-distractor LongMemEval (acc 64.2 vs 62.2) — while outrunning both mem0 and BM25 on both benchmarks. Same answerer, judge, prompt, and budget; only the memory differs.

✨ Why Mneme

• Zero cost. No write-time LLM calls, no embeddings, sub-millisecond recall, storage is plain JSON files.
• Competitive quality. Matches or beats embedding + LLM memory on standard benchmarks (see below).
• Interpretable. Retrieval is explicit cue overlap ranked by IDF — no opaque similarity threshold to tune.
• Human-like forgetting. What has no distinctive cue, or whose cues never recur, is simply never surfaced — forgetting as retrieval failure, for free.
• Local & sovereign. Your memory is a folder of files you can read, edit, or delete. English by default; Chinese optional.

📊 Benchmarks (controlled: same answerer/judge/prompt/budget, only the memory differs)

LoCoMo — low-distractor dialogue (F1 / accuracy):

LongMemEval — high-distractor, ~50 sessions per question (accuracy):

Zero-cost Mneme reaches dense retrieval on LoCoMo and overtakes it where distractors dominate — the regime that matters for real long-term memory — and it is reader-agnostic (near-identical with two different answerer LLMs) and interpretable. The best edge mix is regime-dependent: discourse edges help clean dialogue, associative edges win high-distractor histories.

🚀 Install

pip install mnemekit
python -m spacy download en_core_web_sm     # English model (required)

pip install "mnemekit[zh]"                   # optional: Chinese (jieba)

⚡ Quick start

from mneme import Memory

m = Memory(root="~/.mneme")

m.remember("My dog Lucky is a golden retriever", "Cute! How old is Lucky?",
           session_id="chat-42", round_id=1)

m.recall("what breed is my dog")     # -> [Turn(...), ...]

Turns are stored one file per round at {root}/{session_id}/{YYYY-MM-DD}/{round_id}.json, so any session is trivial to inspect, export, or delete.

🔍 How it works

INGEST (no LLM, no embedding)
  turn ──▶ tag extraction (spaCy NER + lemmas / jieba) ──▶ inverted index  (tag → turns)
                                                           + event tag inheritance

RECALL (per query) — spread over an ephemeral heterogeneous graph, then discard it
  query ─▶ ① lexical: IDF cue reinstatement ─seeds─┬─▶ ② associative edge (shared tag, 1 hop)
                                                   └─▶ ③ discourse edge (adjacent turn in event)
                                                                        │
                                                          recalled context ─▶ your reader

1. Cue index. Each turn is tagged deterministically; tags form an inverted index (a sparse cue index, à la hippocampal indexing). Keyword-free follow-ups inherit their event's tags, so they stay recallable.
2. Reinstatement, then two kinds of edge. A query reinstates directly-cued turns lexically, then spreads activation over an ephemeral graph with two edge types: associative edges (shared discriminative cue → one hop) reach the associative tail a direct cue can't; discourse edges (adjacent turns in the same event) pull in the local conversational context around a hit. Neither is materialized — both are read straight off the index and event structure, used for the one query, and discarded.

🀄 Chinese & custom dictionaries

m = Memory(root="~/.mneme", userdict="my_terms.txt")   # jieba user dictionary
m.remember("我养了只金毛狗，叫 Lucky", "可爱！", session_id="u1", round_id=1)

Chinese is segmented by jieba; a custom dictionary keeps vertical-domain terms intact.

📄 Paper & reproduction

The write-up (method, cognitive grounding, controlled comparison, ablations) is in paper/; the evaluation harness (LoCoMo / LongMemEval / mem0) is in eval/. The frozen reproduction version is on the paper-repro branch.

📌 Citation

@misc{mneme2026,
  title  = {Do Conversational Agents Need Vectors? A Zero-LLM, Zero-Embedding Tag Memory},
  author = {Hao, Shaochun},
  year   = {2026},
  note   = {https://github.com/FTP2026/Mneme}
}

License

MIT