# keymem

[![npm version](https://img.shields.io/npm/v/keymem)](https://www.npmjs.com/package/keymem)
[![Node.js](https://img.shields.io/badge/Node.js-20%2B-green)](https://nodejs.org/)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)

**The associative memory layer for LLM agents — recall by association, not just similarity.**

![keymem chain recall — one recall() per hop walks Key → Memory → Key; skipped branches stay one call away](docs/keymem-chain-recall.gif)

Most agent memory is a vector store. It surfaces what *sounds like* your query — and misses everything your query is *connected to*.

`keymem` stores memories in a **key graph** instead. A search for **"Newton"** can still reach **"strawberries"** — Newton → apple → fruit → strawberry. The path lives in the graph, not in embedding space. It runs locally as an **MCP server**, so any MCP-compatible agent gets human-like associative recall with no external database.

**Works with:** Claude Desktop · Claude Code · any MCP-compatible LLM agent

---

## Why associative memory?

Vector-store memory retrieves by embedding similarity. That works until the thing you need *isn't similar to the words you typed*:

```
Query: "Newton"
Similarity search finds: "Newton discovered gravity"  ✅
Similarity search misses: "user likes strawberries"   ❌
```

A person makes the leap anyway — Newton reminds them of the apple, apples are fruit, they like strawberries. `keymem` makes that same leap because the **path exists in the key graph**: `Newton → apple memory → fruit key → strawberry memory`. No embedding distance connects "Newton" and "strawberry"; a chain of shared keys does.

This is the core idea: memories are not islands ranked by distance. They are nodes in an **N:M key/value graph** that an agent can walk.

---

## How it works

```
Key Space (concepts)         Value Space (memories)
[apple]   ────────┬─────────→      ↑ same memory
[gravity] ────────┘
                  │
[apple]   ────────┼─────────→ "apples are red fruit"
[fruit]   ──────┬─┘
[red]     ──────┤
                │
[fruit]   ──────┼─────────→ "user likes strawberries"
[strawberry]────┘
```

Memories live in a **Value Space**, reached through a separate **Key Space** — one memory reachable via many keys, one key leading to many memories.

`recall("Newton")` returns matching key clusters such as `[Newton]` and `[apple]` plus one passive Top-1 memory under the strongest key. That memory carries `matched_key`, `validity`, and `connected_keys`, so the agent can answer immediately or continue through `read_key(fruit)`.

The default MCP flow remains **Key → Memory → Key**, but its first Key → Memory hop is completed in one call. Only one memory enters context automatically; later hops remain deliberate.

The animation at the top shows this on a real store: three `recall()` calls walk from a profile key to a verification philosophy, loading three memories (~2.4k tokens). The branches the agent skipped are still one call away, and auto-expanding the same 12-memory hub would have cost ~8k tokens. The graph knows the roads; the agent decides the steps. Source: [`docs/demo/chain-walk.html`](docs/demo/chain-walk.html).

---

## Quick Start

> **keymem is an MCP server (a CLI), not a library.** Run it with `npx -y keymem` (recommended —
> always the latest) or install the command globally with `npm i -g keymem`. **Do not** add it to
> your app with `npm i keymem` as a dependency: it bundles `openai`, `zod`, and the MCP SDK, so
> inside an existing project it just duplicates those trees (and can clash with your app's `zod`/
> `openai` versions). The `npm i keymem` line npm shows on the package page is for libraries — it
> doesn't apply here.

```bash
# Optional global install (npx needs none). This puts a `keymem` command on PATH that
# MCP clients can spawn. Run bare, it starts a stdio MCP server and waits for a client —
# so point your MCP config at `keymem` (or just use `npx -y keymem` as shown below).
npm i -g keymem
```

### Claude Desktop

Add to `claude_desktop_config.json`:

**OpenAI embeddings:**
```json
{
  "mcpServers": {
    "keymem": {
      "command": "npx",
      "args": ["-y", "keymem"],
      "env": {
        "OPENAI_API_KEY": "your-openai-api-key"
      }
    }
  }
}
```

**Local embeddings (no API key required) — bge-m3 recommended:**
```json
{
  "mcpServers": {
    "keymem": {
      "command": "npx",
      "args": ["-y", "keymem"],
      "env": {
        "EMBEDDING_BACKEND": "local",
        "LOCAL_EMBEDDING_MODEL": "bge-m3"
      }
    }
  }
}
```

> `bge-m3` (multilingual, recommended) **auto-downloads ~570MB on first run**, then caches. Omit `LOCAL_EMBEDDING_MODEL` for the lighter default (`fast-multilingual-e5-large`). Cross-encoder reranking is part of the core recall path and downloads a second model on first use; set `"KEYMEM_RERANK": "false"` only to disable it.

### Plugin (recommended — Claude Code & Codex)

The repo is also a plugin marketplace, so one install wires up everything: the MCP server
(daemon-backed shim), the `UserPromptSubmit` hook that passively surfaces related memories on every
prompt, and the `keymem` skill carrying the recall/remember protocol.

**Claude Code:**
```
/plugin marketplace add donggyun112/keymem
/plugin install keymem@keymem
```

**Codex CLI:**
```bash
codex plugin marketplace add donggyun112/keymem
codex plugin add keymem@keymem
```

Codex prompts once to trust the hook; approve it or the push path stays silent. The plugin defaults
to local `bge-m3` embeddings (auto-downloads ~570MB on first run, no API key). For OpenAI
embeddings, use the manual setup below instead — plugin MCP servers only see the env they declare.
The synchronous hook waits up to 9 seconds for `/inject` (set `KEYMEM_HOOK_TIMEOUT_MS` to tune it);
local embedding and reranking can add several seconds to prompts that trigger recall.

### Claude Code (manual)

```bash
# OpenAI embeddings
claude mcp add keymem -e OPENAI_API_KEY=your-key -- npx -y keymem

# Local embeddings (no API key required) — bge-m3 recommended (auto-downloads ~570MB on first run)
claude mcp add keymem -e EMBEDDING_BACKEND=local -e LOCAL_EMBEDDING_MODEL=bge-m3 -- npx -y keymem
```

### Codex CLI (manual)

```bash
# OpenAI embeddings
codex mcp add keymem --env OPENAI_API_KEY=your-key -- npx -y -p keymem@latest keymem-shim

# Local embeddings (no API key required)
codex mcp add keymem --env EMBEDDING_BACKEND=local --env LOCAL_EMBEDDING_MODEL=bge-m3 -- npx -y -p keymem@latest keymem-shim
```

Use the `keymem-shim` entry point (not bare `keymem`): it runs the shared daemon the push-path hook
talks to. For the hook, add to `~/.codex/config.toml`:

```toml
[[hooks.UserPromptSubmit]]
[[hooks.UserPromptSubmit.hooks]]
type = "command"
command = "node /absolute/path/to/keymem/hooks/keymem-hook.mjs"
timeout = 10
```

Codex only forwards the env vars declared in its MCP entry, so pass every `KEYMEM_*` /
`SUPER_MEMORY_*` override with `--env`.

That's it — recall and remember work immediately. The agent calls `recall` before its first reply, navigates with `read_key`/`read_memory`, and saves with `remember`.

For reliable proactive saving in Claude Code, add the following to `~/.claude/CLAUDE.md` (MCP prompts are not automatically applied as persistent Claude Code instructions):

```markdown
## keymem

- Before the first reply and whenever the topic changes, call `recall` silently with short noun-keyword queries.
- Before ending every reply, check whether this turn revealed a durable fact: a name, preference, decision, correction, project fact, or goal.
- If it did, call `remember` or `remember_batch` silently in the same turn with 3-6 diverse keys. A durable fact left unsaved is a bug.
- Use `correct` when existing information changes. Save nothing only when the turn revealed nothing durable.
- Treat `read_memory` as retrieval, not confirmation. Use its `validity.status`: qualify `aging`, and never assert `stale` as current without checking an external source or asking the user.
- Call `confirm_memory` only after an explicit current user assertion, an authoritative current source, or direct observation — never merely because a read succeeded.
- Never mention memory lookup or saving to the user.
```

In Codex, put the same block in `~/.codex/AGENTS.md`. (The plugin install ships this as the `keymem`
skill instead, so you can skip it there.)

For other MCP clients, include the `memory_system_prompt` MCP prompt in the agent's persistent system instructions. It mandates 3 blind recall() calls before the first reply — appropriate when there's no hook surfacing anything passively. If you *do* wire up a `UserPromptSubmit`-equivalent push path yourself, set `KEYMEM_HOOK_INSTALLED=true` so the prompt reacts to what was surfaced instead of blind-guessing on top of it (this is set automatically by the plugin's `.mcp.json`).

### Manual / Development

```bash
git clone https://github.com/donggyun112/keymem
cd keymem
pnpm install
```

Create `.env`:
```
OPENAI_API_KEY=your-openai-api-key
OPENAI_EMBEDDING_MODEL=text-embedding-3-small
```

Or use local embeddings (no API key required):
```
EMBEDDING_BACKEND=local
LOCAL_EMBEDDING_MODEL=fast-multilingual-e5-large  # default; best fit for Korean/multilingual keys
```

```bash
pnpm dev
# or:
pnpm build
pnpm start
```

**Requirements:**
- Node.js 20+
- pnpm for local development
- OpenAI API key for OpenAI embeddings, or `fastembed` for local embeddings

---

## Features

- **N:M key/value graph** — memories and the concepts that index them are separate spaces, linked many-to-many. One memory is reachable through many keys; one key leads to many memories.
- **Agent-driven Key → Memory → Key navigation** — the agent walks the graph deliberately instead of collapsing it into one opaque similarity search.
- **Associative multi-hop recall** — reach memories no embedding distance would connect, by following chains of shared keys.
- **Confirmation-aware freshness** — reads learn useful paths without certifying old content; explicit evidence refreshes facts across four decay profiles.
- **One-step versioning** — a correction retains its immediate predecessor and records what it superseded.
- **Key types** — `concept` keys match by similarity; `name`/`proper_noun` keys match exactly, so "동건" never matches "뉴턴" just for being short.
- **Cross-lingual key merging (IDF)** — `파이썬` and `Python` collapse into one canonical cluster instead of fragmenting the key space.
- **Hebbian link learning** — the path an agent actually traverses gets reinforced ("fire together, wire together"), so useful associations become easier to reach.
- **Hybrid retrieval** (optional direct mode) — Kiwi BM25 over keys and content + key embeddings + Reciprocal Rank Fusion, with depth/confirmation-freshness modulation and configurable multi-hop expansion.
- **Cross-encoder reranking** (core default) — `bge-reranker-v2-m3` re-scores the passive Top-1 pool and compatibility direct-mode candidates.
- **Local-first** — all data in a local JSON graph; no external database. OpenAI or fully-local embeddings (auto-downloaded).

### Freshness and Depth

Every memory has a depth score `0.0 → 1.0`, but retrieval and confirmation are separate:

| Stage | Depth | Behavior |
| --- | --- | --- |
| Shallow | `< 0.3` | Little explicit confirmation; minimal ranking boost. |
| Medium | `0.3–0.7` | Repeatedly confirmed; moderate ranking boost. |
| Deep | `> 0.7` | Strongly confirmed; maximum ranking boost, but still correctable and age-sensitive. |

`read_memory()` increments access metadata and may reinforce the traversed key edge, but it does
**not** change depth, `last_confirmed_at`, `confirmation_count`, or freshness. Only
`confirm_memory(memory_id, evidence)` refreshes validity and increases depth `+0.05`; accepted
evidence is an explicit current user assertion, an authoritative current source, or direct
observation. A successful read alone is never confirmation.

Freshness decays from `last_confirmed_at` according to the memory's `decay_profile`:

| Profile | Default half-life | Intended use |
| --- | --- | --- |
| `transient` | 7 days | Fast-changing state such as temporary plans or availability |
| `standard` | 90 days | General facts; the default |
| `stable` | 365 days | Slowly changing facts |
| `permanent` | No decay | Deliberately timeless or immutable facts |

Every memory view includes a `validity` payload with `freshness`, `status` (`fresh`, `aging`, or
`stale`), `age_days`, `last_confirmed_at`, `confirmation_count`, `decay_profile`,
`verification_recommended`, and `verification_required`. `fresh` means freshness is at least
`0.5`; `aging` is at least `0.125` but below `0.5`; `stale` is below `0.125` and must not be
asserted as current without verification. Decay is soft: it lowers ranking through
`0.2 + 0.8 × freshness` and never deletes a memory. TTL remains the only automatic expiry mechanism.

For example, save a temporary plan with `remember(..., decay_profile:"transient")`; after the user
explicitly says it is still current, call `confirm_memory(memory_id, evidence:"user")`. Do not
confirm merely because `read_memory` returned it.

### Key Types

Not all keys should behave the same. Names shouldn't match semantically — "동건" shouldn't match "뉴턴" just because they're both short Korean words.

| Type | Matching | Use Case |
| --- | --- | --- |
| `concept` (default) | Embedding similarity ≥ threshold (0.28 OpenAI / 0.60 local) | Topics, categories, attributes |
| `name` | Exact match only | Person names |
| `proper_noun` | Exact match only | Brands, places |

Name/proper_noun keys also get an IDF penalty (`×0.5`) when they become hub keys connected to many memories, preventing them from polluting unrelated searches.

### Versioning (one-step predecessor)

```
"user lives in Seoul"   (depth: 0.4 → weakened to 0.12, preserved)
        ↑ superseded by
"user moved to Busan"   (depth: 0.0, new)
```

`keymem` retains the immediate predecessor instead of overwriting it. The superseded record is
excluded from active retrieval regardless of its depth, and the new version becomes current. A
later correction prunes the grandparent, so provenance is one step rather than a full-history
archive.

### Key Merging

```
Add key "파이썬"  → finds existing "Python" (similarity 0.87 > threshold 0.85)
                 → reuses existing key instead of creating duplicate
```

Prevents key space fragmentation. The same concept across languages or phrasing stays unified.

### Agent-driven Retrieval (default)

The default MCP API crosses from Key Space to one Value while preserving explicit graph navigation:

1. `recall(query, context)` returns ranked key clusters and the Top-1 memory under the strongest key. The raw `context` ranks memories within that key.
2. The memory is a passive preview with `validity`, `matched_key`, and every `connected_keys[].key_id`. It does not increment access/depth, reinforce links, learn aliases, or confirm freshness.
3. The agent checks which facts the question needs. It can use the preview when those facts are present; otherwise it follows a connected key for a missing fact with `read_key(key_id, missing_fact_query)` or recalls that fact directly. It stops when the evidence is complete or no promising path remains. Later memories are not automatically injected.
4. `read_memory(memory_id, via_key_id)` remains the explicit full-read path. It updates access metadata and reinforces only the traversed edge; it does not confirm or deepen the memory.

Semantically merged keys are preserved as aliases on one canonical key cluster (for example `Python` + `파이썬`). The recommended `bge-m3` profile enables conservative short-key merging by default; override or disable it with `KEYMEM_SHORT_KEY_MERGE`. A key linked to at least three active memories is surfaced as a hub with `is_hub`, `memory_count`, and `specificity` metadata rather than being hidden by IDF. Override the hub threshold with `KEYMEM_KEY_HUB_MIN_LINKS`.

### Associative Recall Engine (library API, used internally by `recallInject`)

`MemoryGraph.recall()` is a one-call ranked retrieval path used by `recallInject()` and available to library callers. It combines two candidate paths with **Reciprocal Rank Fusion** (`RRF_K = 60`):

- **Kiwi BM25:** lexical search over memory keys and content, using the official `kiwi-nlp` WASM tokenizer in Node. Korean morphology is tokenized locally; no Python process is required.
- **Key matching:** query embeddings match concept keys; exact names and aliases provide literal entry points. Key links lead to candidate memories.

Memory bodies have no embeddings. Cross-encoder reranking remains enabled and scores query–content pairs directly. Graph expansion, depth, and confirmation freshness still affect ranking. The default MCP `recall` tool uses explicit key navigation, so agents can drive later hops deliberately.

### Hebbian Link Learning

Reading a full memory is a **write**, not just a read. In the default flow, `recall()` and `read_key()` are read-only; `read_memory(memory_id, via_key_id)` reshapes the selected path:

- The traversed `via_key_id → memory_id` link is **reinforced** (`+0.1`, capped at `3.0`).
- A full read increments access metadata but changes neither depth nor `last_confirmed_at`, so it
  never refreshes freshness. Freshness continues to decay as elapsed time grows. Evidence-backed
  `confirm_memory()` increases depth and refreshes freshness by advancing `last_confirmed_at`.

Reinforcement is scoped to the key the agent actually traversed — not every key attached to the memory. This is the literal Hebbian rule ("fire together, wire together") and prevents unrelated associations from growing when the memory is reached through a different concept. Weights are clamped to `[0.1, 3.0]`.

Link weights feed back into `read_key()` ranking, so repeatedly selected paths become easier to reach. `recall()`/`recallInject()` retain the previous matched-link reinforcement and explored-link decay behavior.

---

## Architecture

```
│                      Key Space                          │
│   [name] [동건] [programming] [python] [fruit] [red]   │
│      ↓      ↓         ↓           ↓       ↓      ↓     │
│   [vec]  [exact]    [vec]       [vec]   [vec]  [vec]   │
                         │ N:M links
                         ↓
│                     Value Space                         │
│   "user's name is Donggeon"     depth: 0.85  (deep)    │
│   "user likes Python"           depth: 0.30  (medium)  │
│   "user likes strawberries"     depth: 0.05  (shallow) │
```

**Default MCP navigation:**

1. Embed the query and match canonical key concepts plus exact aliases.
2. Return key clusters with `memory_count`, `is_hub`, and `specificity`; do not return memory content.
3. Rank a selected key's memory handles by link weight, depth, and confirmation freshness in `read_key()`.
4. Return full content and adjacent key clusters from `read_memory()`; reinforce the traversed edge only when `via_key_id` is supplied.
5. Repeat `read_key(next_key_id)` to walk the graph deliberately.

**`recall()` algorithm (hybrid, configurable 1–5 hops; default 2; used internally by `recallInject()`):**

1. Tokenize the query and memory keys/content with Kiwi, then retrieve BM25 candidates.
2. Match concept keys with query embeddings and literal name/alias matches, then follow their memory links.
3. Fuse lexical and key-path ranks with RRF (`RRF_K = 60`).
4. Apply depth and confirmation freshness, and expand through shared keys and explicit `related_to` links for the configured hops.
5. Rerank candidates with the cross-encoder and return ranked results. Matched-link reinforcement and explored-link decay retain their existing behavior.

### Storage and duplicate handling

Only keys retain embedding vectors. Existing memory-body vectors are ignored on load and omitted after the next successful save; key vectors, memory text, links, and history are retained. Back up the store before upgrading if you need to reopen it with an older release.

Duplicate detection compares normalized memory text exactly, within the existing namespace and active-memory rules. Paraphrases are separate records. Automatic confirmation requires normalized utterance and body text to match exactly. New memories do not gain links or contradiction flags from body-vector similarity; supply keys and explicit relationships, and use `correct` for changed facts. Existing explicit links and recorded contradictions remain readable.

### Key similarity thresholds

Embedding thresholds apply to keys, not memory bodies. `getThresholdProfile()` in `src/embedding.ts` supplies model-specific defaults; use `KEYMEM_KEY_RECALL`, `KEYMEM_KEY_MERGE`, `KEYMEM_KEY_AUTOLINK` (auto-key alias promotion), and `KEYMEM_SHORT_KEY_MERGE` to tune key routing and canonicalization. An uncalibrated model falls back to the BGE profile with a warning.

Legacy body-cosine settings (`KEYMEM_CONTENT_RECALL`, `KEYMEM_CONTENT_RECALL_SHORT`, `KEYMEM_MEMORY_DEDUP`, `KEYMEM_CONTRADICTION`, `KEYMEM_MIN_SCORE`, `KEYMEM_GATE_Z`, and `KEYMEM_KEY_GATE`) no longer control memory retrieval or writes. BM25 scores, key cosine scores, RRF ranks, and cross-encoder scores are different scales; do not reuse an old body-cosine threshold as a lexical relevance threshold.

| Env var | Default | Description |
| --- | --- | --- |
| `KEYMEM_AUTOKEY` | `true` | Auto-key self-healing: learn missing search terms from real usage. Set `false` to disable. |
| `KEYMEM_AUTOKEY_PROMOTE_N` | `3` | Routing-confirmed selections of a `(key, query)` pair before the query is folded into the key space. |
| `KEYMEM_AUTOKEY_CONFIRM_FLOOR` | `0.45` | Lowest query↔key cosine eligible for routing-confirmation learning. Repeated selections through the same key can teach a below-gate query alias; this confirms routing only, never content freshness. Lower (e.g. `0.40`) to catch more borderline paraphrases; set `≥` the recall threshold to disable. |
| `KEYMEM_AUTOKEY_MAX_ALIASES` | `8` | Max learned aliases promoted per key. |
| `KEYMEM_AUTOKEY_PRUNE_AGE` | `2592000` | Seconds before a never-hit learned alias is pruned by the daemon's automatic TTL sweep (30 days). |
| `KEYMEM_CLEANUP_INTERVAL_MS` | `3600000` (1h) | How often the daemon sweeps expired memories/aliases on its own. The stdio path sweeps once at startup instead (no long-running loop to schedule against). |
| `KEYMEM_DECAY_TRANSIENT_DAYS` | `7` | Half-life in days for `transient` memories. Must be finite and greater than zero. |
| `KEYMEM_DECAY_STANDARD_DAYS` | `90` | Half-life in days for the default `standard` profile. Must be finite and greater than zero. |
| `KEYMEM_DECAY_STABLE_DAYS` | `365` | Half-life in days for `stable` memories. Must be finite and greater than zero. |

For cross-language recall, retain multilingual keys and aliases (for example `딸기` and `strawberry`). Lexical body search alone does not translate queries.

---

## MCP Tools

The tool set contains 10 tools:

| Tool | Description |
| --- | --- |
| `recall(query, top_k?, namespace?, context?, explain?)` | Return ranked keys plus one passive Top-1 memory with `validity`, `matched_key`, and `connected_keys`. `explain:true` also returns namespace counts. |
| `browse_keys(namespace, hubs_only?, limit?, offset?)` | Browse a namespace's active key vocabulary, hubs first, when recall has no entry hit. |
| `read_key(key_id, query?, namespace?, limit?, offset?)` | List ranked memory IDs and metadata connected to one key. Pass the original query for relevance ordering; supports pagination for hubs. |
| `read_memory(memory_id, via_key_id?, namespace?)` | Read full memory content, connected keys, and `validity`. Updates access and, when `via_key_id` is supplied, that traversed edge; never confirms or deepens content. |
| `confirm_memory(memory_id, evidence, namespace?, source?)` | Refresh freshness and deepen a current memory after explicit user evidence, an authoritative source, or direct observation. A read alone is not evidence. |
| `remember(content, keys, key_types?, namespace?, ttl_seconds?, decay_profile?, related_to?)` | Save memory with key concepts, optional TTL, and a `transient`, `standard`, `stable`, or `permanent` decay profile. |
| `correct(memory_id, content, keys?, key_types?, ttl_seconds?, decay_profile?, related_to?)` | Versioned update. The immediate predecessor is preserved but inactive; omitted TTL/profile inherit from it. |
| `dismiss(memory_id, key_id, namespace?)` | Negative feedback: the fact is fine, this key should not have surfaced it. Weakens that one edge (floored, never severed) and cancels its pending alias learning |
| `forget(memory_id)` | Permanently delete |
| `remember_batch(items)` | Save multiple memories; each item accepts `ttl_seconds` and `decay_profile` |

TTL-expired memories and key/memory/link counts are no longer agent-facing tools: the daemon
sweeps expired memories itself (once at startup, then every `KEYMEM_CLEANUP_INTERVAL_MS`; the
stdio path sweeps once at startup), and the key/memory/link stats are already passively injected
into `SERVER_INSTRUCTIONS` on every turn — a separate `memory_stats()` call was pure duplication.

Scores carry `score_kind`. Key recall exposes `key_relevance` from key cosine or bounded lexical relevance. `content_relevance` is bounded BM25 (`s / (1 + s)`), not cosine. `read_key` uses `within_key_rank`, with cross-encoder relevance when available and bounded lexical relevance otherwise. Direct recall uses `rrf_fused_rank`; RRF `rank_score` is a rank-fusion value, not a similarity probability.

A system prompt template is also available via the `memory_system_prompt` MCP prompt — include it to instruct the agent to recall silently, use diverse keys, and never mention the memory system to users.

---

## Local embedding models

**BGE-M3 (recommended, multilingual) — auto-downloaded:** set `LOCAL_EMBEDDING_MODEL=bge-m3` (aliases: `bgem3`, `baai/bge-m3`, `fast-bge-m3`). On first use the model is **fetched automatically if missing** — quantized ONNX (~570MB, from `onnx-community/bge-m3-ONNX`) plus the tokenizer/config (from `BAAI/bge-m3`) — and cached under `~/.keymem/models/bge-m3`. No manual download needed:

```
EMBEDDING_BACKEND=local
LOCAL_EMBEDDING_MODEL=bge-m3
# optional — point at an existing model dir to skip the download (backward compatible):
# LOCAL_EMBEDDING_MODEL_PATH=/absolute/path/to/model-dir   # dir with model.onnx + tokenizer files
# LOCAL_EMBEDDING_MODEL_FILE=model.onnx                    # optional; default is model.onnx
# optional: KEYMEM_EMBED_THREADS=4                         # ONNX intra-op threads (default: a quarter of the machine, max 6)
```

> First run downloads ~570MB once, then reuses the cache. If `LOCAL_EMBEDDING_MODEL_PATH` already holds the model it is used **as-is with no download** (a partial dir is self-healed — only missing files are fetched). Online-API backends (OpenAI) and fastembed built-ins are unaffected.

> **bge-m3 runs on its own ONNX session, not through fastembed.** Inputs use their actual token length. Its vector-space fingerprint is `local:bge-m3+nopad`; a fingerprint change triggers key-vector migration. Earlier padding/latency and retrieval measurements in [BENCHMARKS.md](BENCHMARKS.md) describe the previous body-embedding engine.

**Cross-encoder reranking (core):** the default `recall` MCP tool's Top-1 path re-scores the candidate pool under its strongest key with `bge-reranker-v2-m3`. `recall()`/`recallInject()` results are reranked too. The model (~570MB, quantized) auto-downloads on first use and caches under `~/.keymem/models/reranker`.

```
# optional: KEYMEM_RERANK=false              # disable the core reranker
# optional: KEYMEM_RERANK_MODEL_PATH=/dir    # use an existing model directory
# optional: KEYMEM_RERANK_POOL=30            # candidates re-scored (default 30)
# optional: KEYMEM_RERANK_THREADS=4          # ONNX intra-op threads (default: a quarter of the machine, max 6)
```

> On by default. If the model cannot load, recall falls back to fused ranking. Query decomposition remains the caller's responsibility.

**Task-conditioned evidence selection:** expanded recall stays associative, but comparison-shaped
queries are projected at the final evidence boundary so both named entities receive task-relevant
support. KeyMem derives the comparison dimension after removing the entity names (for example,
`nationality` or `founded`), keeps the graph candidate pool and association scores intact, and
preserves the original associative winner for reinforcement. This is on by default; disable it with
`KEYMEM_TASK_EVIDENCE_SELECTION=false`.

**Reranker not-found gate (`KEYMEM_RERANK_MIN_SCORE`):** in `recall()`/`recallInject()`, reject the complete result when the top cross-encoder logit is below this floor.

```
KEYMEM_RERANK_MIN_SCORE=0   # reject when top rerank logit < 0 (bge-reranker-v2-m3 scale)
```

> ⚠️ **Caveats.** (1) Unset by default — no gate. (2) The logit scale is **model-dependent**; `0` (≈ sigmoid 0.5) suits `bge-reranker-v2-m3` (measured: same-language found ≈ +3.9, not-found ≈ −5 to −6) — recalibrate for other rerankers. (3) **Trusted for SAME-LANGUAGE only.** Cross-lingual relevance logits run low even when relevant (KR query ↔ EN memory ≈ −5.4), so the gate auto-**bypasses on a script mismatch** (KR↔Latin) to avoid false-rejecting cross-lingual hits — which means **cross-lingual content must be reachable via bilingual keys** (`["Jiwoo","지우"]`), and cross-lingual *not-found* precision is a known limitation. Leave this off if you can't tag bilingual keys.

> **Prefix behavior:** BGE-M3 does **not** use `passage:`/`query:` prefixes — embeddings are passed through as-is. All other local models (e5, BGE-en, MiniLM) continue to use prefixes unchanged.

> **Multilingual key embeddings:** `bge-m3` remains supported. Historical model comparisons do not establish the quality or not-found behavior of the current Kiwi BM25 architecture.

If `OPENAI_API_KEY` is not set and `EMBEDDING_BACKEND` is unset, the server automatically uses the local `fastembed` backend.
For English-only use or lower local resource usage, set `LOCAL_EMBEDDING_MODEL=fast-bge-base-en-v1.5` or `fast-bge-small-en-v1.5`.

> **Switching backends is safe.** The graph records an embedding **fingerprint** (backend + model id) identifying the vector space its embeddings live in. On startup, if the current backend's fingerprint **or** dimension differs from what is stored, the graph **auto-migrates** — every key is re-embedded with the new backend while content, links, depth, and access history are preserved (a `graph.json.bak.*` backup is written first). The fingerprint matters because two models can share a dimension yet produce incompatible vectors (e.g. `fast-multilingual-e5-large` and `bge-m3` are **both 1024-d**); a dimension check alone would miss that swap and silently corrupt every similarity. Disable with `KEYMEM_AUTO_MIGRATE=false`. Re-embedding via OpenAI incurs one-time API cost proportional to your key count.
>
> **Migrating a pre-fingerprint (legacy) graph across same-dimension models.** A graph written before fingerprinting has no recorded vector space, so a same-dimension model swap off it cannot be detected automatically. Set `KEYMEM_FORCE_REEMBED=true` for **one** startup to re-embed unconditionally and stamp the fingerprint; remove it afterward (left on, it re-embeds on every start). This is exactly the one-shot needed when moving an existing e5 graph to bge-m3.

---

## Data Storage

All data is local. No external database required.

```
~/.keymem/
└── graph.json              # canonical keys, aliases, memories, weighted links
```

Set `KEYMEM_DATA_DIR` to use a different storage directory.

**Linking a memory to its source conversation.** When you save a memory, keymem stamps the active host session onto its `source` (`host_session` / `host_agent` / `host_turn`) for provenance. keymem does not expose a tool to read those transcripts back — it only reads them internally to resolve the active session. Locations are auto-detected per OS and honour the agents' env overrides:

- **Claude Code** — `~/.claude/projects/**/{session_id}.jsonl` (`$CLAUDE_CONFIG_DIR`)
- **Codex** — `~/.codex/sessions/**/rollout-*-{session_id}.jsonl` (`$CODEX_HOME`)

**Access is gated.** Because transcripts are local, potentially sensitive history, keymem only reads them to stamp provenance **when trusted as the owner's personal local agent** — i.e. a recognized host injected its session env (`CLAUDE_CODE_SESSION_ID` / `CODEX_THREAD_ID`), or you explicitly opt in with `KEYMEM_TRANSCRIPT_ACCESS=true`. Otherwise (a plain server, a remote deployment, a non-owner/custom agent) memories are saved without a host link. Set `KEYMEM_TRANSCRIPT_ACCESS=false` to force-disable even under a host agent.

The active session is found two ways:

1. **Deterministic** — the host injects its session id into every MCP server it spawns, and keymem reads it directly: Claude Code → `CLAUDE_CODE_SESSION_ID`, Codex → `CODEX_THREAD_ID` (which equals the rollout file's session id). The link is exact, with no guessing.
2. **Heuristic fallback** — for hosts that don't expose a session id (e.g. Claude Desktop), keymem uses the most-recently-modified transcript (with a staleness guard).

---

## Benchmarks

Historical measurements of the previous body-embedding engine: on **HotpotQA** bridge questions (real external multi-hop data, gold labels, no LLM judge — and
keys generated *blind* by independent subagents that never saw the question or answer), the
key-graph retrieves **both** gold supporting paragraphs **63%** of the time vs **53%** for flat
semantic retrieval and **35%** for lexical (+10pp / +28pp) — the connected-but-dissimilar case
it's built for. (Honest: my own hand-derived keys inflated this to 78/60; on non-multi-hop
"comparison" questions the graph slightly *hurts*; it's retrieval-recall, not answer accuracy.)
The read path is also O(1) (`read_memory` p50 ~45ms → ~0.01ms @ 500 memories). Full methodology
and caveats: **[BENCHMARKS.md](BENCHMARKS.md)**. These scores do not validate the current Kiwi BM25 + key-embedding engine; no production quality scores have been established for it.

---

## Limitations

- **Linear scan** — suitable for personal use (~10k memories). FAISS/ChromaDB integration planned for larger scale.
- **Later-hop round trips** — `recall` completes the first Key → Memory hop, but following a connected key to another memory still requires `read_key → read_memory`.
- **Hub breadth** — broad keys can connect many memories. `read_key()` paginates hubs; the agent must choose whether to continue paging or follow a more specific adjacent key.
- **Agent quality matters** — key selection on `remember` affects retrieval quality. System prompt tuning is important.
- **Lexical coverage** — Kiwi BM25 does not translate or detect paraphrases by itself. Multilingual keys, key embeddings, and reranking help route and rank candidates, but cannot guarantee missing candidates are recovered.
- **Threshold calibration** — thresholds are tuned per embedding model. A new/uncalibrated model falls back to the BGE profile (with a warning); recalibrate via the `KEYMEM_*` env overrides.

---

## Testing

```bash
pnpm test                        # unit tests (fast, no model download)
tsx test/scenarios.ts            # 21 end-to-end behavioral checks (local e5)
tsx test/robustness.ts           # threshold overrides + Hebbian pollution bounds
tsx test/migration.ts            # backend/dimension switch auto-migration (no brick)
tsx test/nhop.ts                 # N-hop chained traversal (recall hops parameter)
tsx test/depth-noise.ts          # deep-hop noise bounds + relative score floor
tsx test/live-multilingual.ts    # interactive multilingual recall demo

# Manual retriever-quality check (NOT part of pnpm test):
EMBEDDING_BACKEND=local npx tsx test/retriever-quality.live.ts
# For bge-m3: also set LOCAL_EMBEDDING_MODEL=bge-m3 LOCAL_EMBEDDING_MODEL_PATH=/abs/dir
```

`scenarios.ts` and `robustness.ts` exercise the real local embedding backend (direct/associative/cross-lingual recall, versioning, depth growth, dedup, TTL, Hebbian learning, namespace isolation). Key-threshold recalibration and retrieval quality require fresh runs after an engine or model change; historical passing results do not validate this architecture.

---

## Roadmap

- [ ] FAISS/ChromaDB for scale
- [ ] Coding agent profile (different key strategies for code context)
- [ ] Memory export/import
- [ ] Multi-user support

---

## Author

**donggyun112** — [github.com/donggyun112](https://github.com/donggyun112)

Repository: [donggyun112/keymem](https://github.com/donggyun112/keymem) · Issues & PRs welcome.

## License

MIT © [donggyun112](https://github.com/donggyun112)