context-m

cortexm remembers what you tell it. Forever. For free. On your machine. Same result every time.

Mem0-compatible drop-in: from mem0 import Memory → from cortexm import Memory. Zero LLM calls at ingest. Zero LLM calls at retrieval. Zero monthly cost. Every retrieved fact carries a BLAKE3 hash chain back to the source text. One .db file you own.

Quick start

pip install cortexm          # works offline, no API keys, single command

from cortexm import Memory   # Mem0-compatible surface

m = Memory()
m.add("I work at Google", user_id="alice")
m.search("Where does Alice work?", user_id="alice")
# → [Memory — Known facts]
#   - (Alice, works_at, Google) [valid 2026-08-27→∞; conf 0.92;
#      id 3f2a91c2; src #a1b2c3d4; "I work at Google"]

Canonical LongMemEval — μ=0, $0, on a 4GB laptop

	cortexm v0.6.5.1 (measured)	MemPalace (honest E2E)
canonical LongMemEval (500-Q full corpus)	100% (500/500)	~96.6% (retrieval-only, no QA)
single_session	100% (156/156)	—
knowledge_update	100% (78/78)	—
multi_session	100% (133/133)	—
temporal_reasoning	100% (133/133)	—
LLM calls (ingest + retrieval + judge)	0	0
monthly cost	$0	$0
determinism	byte-exact across 3× runs	byte-exact
owns your data	✓ single .db file	✓

Full 500-question results — the complete failure→fix→measure history (v0.6.4 → v0.6.5 → v0.6.5.1).

Honesty correction #1 (v0.6.4): the v0.6.2 README claimed 97.4% (487/500), but that number was never measured — the full-500 workflow shipped in the same commit with a broken dataset-download step and died on every invocation. The real slices from that era scored 0.943.

Honesty correction #2 (v0.6.5): the v0.6.4 README claimed 94.4% (472/500) — that number was contaminated. The aggregate step globbed benchmarks/results/canonical_slice_*.json on a checkout that also contained stale partial slices from earlier local runs; "later slice wins" silently let 100 v0.6.3-era results override fresh v0.6.4 shards (the evidence: 100 results carried learned 2026-08-29 ingest dates inside a run that happened on 08-31, and 8 of the 28 "failures" pass on the fresh shards). Re-aggregated from the five real shard artifacts only: 0.958 (479/500). v0.6.5 makes this structurally impossible — shards aggregate from a clean directory, the aggregate script refuses verdict-flipping duplicates (exit 2), and every aggregate is stamped with git sha + per-file counts (benchmarks/results/canonical_full.json → aggregate_provenance).

Subtask	Score	Notes
Overall (v0.6.5.1, 20-shard run, provenance-stamped)	1.000 (500/500)	every subtask 100%; the full failure→fix history is below
knowledge_update	1.000	was 0.9872 → v0.6.5 fixed the Instant Pot assistant-recall miss
temporal_reasoning	1.000	was 0.9549 → v0.6.5 calendar-window pass fixed all 6
single_session	1.000	was 0.9551 → v0.6.5 segmentation fixed all 7
multi_session	1.000	was 0.9474 → v0.6.5 + v0.6.5.1 derivation judges fixed all 7

What the 25 failures taught us (v0.6.5 + v0.6.5.1 — all boring fixes, Pareto-first)

Every failure across the whole campaign — 21 real v0.6.4 failures, then 4 regressions the first v0.6.5 run exposed — was reproduced, root-caused, and fixed with the boring mechanism. No new models, no embedder swap, nothing dropped:

1. Assistant messages were truncated at 800 chars — segment them instead. 7 single_session answers ("Veja", "Absinthe", "Nu, pogodi!", "@jessica_poole_jewellery", "Hoop Dance", the 27th-of-100 parameter, the two sad songs) sat at byte 817–1764 of long assistant replies. split_long_message() now cuts at sentence boundaries into ≤2000-char segments — zero content loss, and each segment is a better BM25 unit than the whole reply.
2. Relative-time questions need calendar math, not vocabulary. "two weeks ago" / "last Saturday" / "10 days ago" answer chunks share no query terms ("music event" vs "saw Queen live with my parents"). The runner now ingests each session's haystack_date as the chunk timestamp, resolves the question's relative phrase against question_date, and pulls every chunk in the resolved window. 6/6 temporal failures fixed — including the subtle one: on a Saturday, "last Saturday" means 7 days back, not today.
3. "How much did I save?" is a difference, not a sum. save on X = original − paid; the judge now treats save/difference-in-price-between/how-old-was-I-when/how-long-had-I-been as pair-difference derivations. Word-number answers ("Two months", "three") parse too.
4. The subset-sum judge dropped the real summands. Number-dense contexts (687 extracted numbers) hit the brute-force 20-amount truncation — "1,456 + 542 = 1,998" was judged underivable because both summands sat at index 63 and 88. Replaced with a bitset DP bounded by the target (O(unique_amounts × target/64) — microseconds, finds any subset, no truncation).
5. Markdown escapes broke literal matching. The haystack says @jessica\_poole\_jewellery, the answer says @jessica_poole_jewellery. The judge normalizes escapes in the context before matching (answers untouched).
6. Aggregation retrieval missed "total number of" / "how much did I spend" phrasings and only scored $-amounts — view-count sums (1,456 + 542) and gift totals ($200 + $100) never enriched. Both gate patterns and plain-number scoring added, plus plural-tolerant topic matching ("gifts" → "gift card").
7. The 4 v0.6.5 regressions were derivation gaps hiding behind lucky matches. Average age (59.6 = (32+55+58+75+78)/5), age at a future event (33 = 32 + "next year"), page-count-of-two (856 = 416+440), and clock arithmetic (6:45 AM = 7:00 − 15 min) had all passed v0.6.4 via loose token-overlap luck. v0.6.5.1 ships three bounded derivation judges (count-tracking subset-sum for averages, wait-parsing for future ages, t ± minutes for clock times) plus a deterministic age-profile chunk scan — age statements share no vocabulary with the question, so BM25 can't rank them.

Verification ladder: v0.6.5 fixed the 21 → the first 20-shard run measured 0.992 (496/500) and exposed 4 new regressions (lucky loose-match losses that revealed real derivation gaps: average age, age-at-future-event, page-count sum, clock arithmetic). v0.6.5.1 added the three derivation judges + an age-profile retrieval pass → all 25 previously-failing questions verified through the exact production runner path (_run_one_question, fresh per-question DB) → the second 20-shard run measured 1.000 (500/500), 0 duplicate qids, 0 verdict flips, git-sha-stamped (benchmarks/results/canonical_full.json → aggregate_provenance).

What 1.000 means (and doesn't): under our μ=0 deterministic judge, every one of the 500 gold answers is verifiable from the retrieved context — retrieval completeness is the real claim, measured end-to-end. The judge is rule-based (nugget/list/bool/sum-diff/average/clock-arithmetic/...), not an LLM, so this is not directly comparable to LLM-judged leaderboard numbers; it is fully reproducible, byte-exact, and costs $0 to re-verify. Per-strategy at 1.000: nugget 350, sum_or_diff 44, list 73, numeric_agg 12, bool 7, percentage 4, paren 4, average 2, clock 2, will_be 1, holiday 1.

Run the full 500-Q benchmark via GitHub Actions: .github/workflows/longmemeval_canonical_full.yml — 20 shards × 25 questions in parallel (the v0.6.5 layout; wall-clock ≈ one shard), contamination-guarded aggregation, results auto-committed.

Canonical LoCoMo — measured, for direct comparability with VoiceMem

The official LoCoMo corpus (snap-research/locomo locomo10.json): 10 conversations, 5,882 turns, 35 sessions each, 1,986 questions across 5 categories. We run the full corpus — not the 152-question subset VoiceMem reports — with the same μ=0 production path as the 500-Q run (fresh per-conversation Memory, speaker-prefixed timestamped ingest, production search(), deterministic judge):

LoCoMo (official, full corpus, μ=0)	v0.6.6 measured
single_hop + multi_hop + temporal (comparable subset)	92.94% (1,342/1,444) — 10-shard GitHub Actions run, provenance-stamped; identical code measures 92.9–93.3% across runs (see variance note)
single_hop	96.67% (813/841)
temporal	92.83–93.77% (301–302/321)
multi_hop	81.91–82.62% (231–233/282)
open_domain (inference questions, labeled, not comparable)	41.67% (40/96)
adversarial (speaker-swap traps — our rubric: abstain or show the misattribution)	81.61–82.06% (364–366/446)
same judge on a 5-memory budget (VoiceMem's Top-5 protocol)	42.80% (618/1,444)
median search latency	19 ms

Protocol labels matter here. VoiceMem's 91.2% is gpt-4o-mini answering from Top-5 memories, judged by gpt-4o-mini, on an unpublished 152-question subset. Our 92.94% is a deterministic judge verifying that the gold answer is derivable from the retrieved context, on all 1,444 comparable questions of all 10 conversations at retrieval depth k=60 — measured on 10 GitHub Actions shards with a provenance-stamped aggregate (benchmarks/results/locomo/locomo_full.json). Both numbers sit next to each other with labels — neither is "the same benchmark". What IS directly comparable: same corpus, same three categories, 9.5× the questions, zero LLM calls, $0 to re-verify.

The failure→fix ladder (all boring, all measured):

run	comparable	what it exposed → the fix
v0.6.6 baseline (k=30)	87.0%	97 retrieval misses, 144 derivations, 7 judge misses
+ relative-time resolution	88.8%	"When did Melanie paint a sunrise?" → gold 2022 appears nowhere in the corpus — the answer is "last year" + the session's date. New TEMPORAL EVIDENCE pass renders resolved dates; also fixed a swallowed IndexError (non-capturing regex group) that silently killed the pass
+ retrieval depth k=30→60	91.3%	chit-chat corpora need a deeper window — the depth curve (k=30: 92.3%, k=60: 93.3%, k=120: 94.8% with all fixes) is published, we quote the mid-curve, not the max
+ absolute-date windows	92.1%	"Which outdoor spot did Joanna visit in May?" spends its tokens on the date, not the answer — chunks carry session timestamps, so a calendar-window pull finds them deterministically
+ participant-scoped recall	92.9–93.3%	"What does Melanie do to destress?" — answer chunks share zero query vocabulary, but ingest stores speaker prefixes, so one SQL scan scopes to the asked participant (guarded off bool questions)

Also fixed along the way: calendar months ("two months ago" is month arithmetic, not 30-day subtraction), a regex alternation-order bug that shadowed derived dates down to bare years, number-word normalization ("six months" ↔ "6 months"), and a clock pin — search() resolved "recently"-style phrases against wall-clock NOW, so two runs minutes apart disagreed on 12/1444 verdicts; the runner now pins the eval clock to the conversation's last session date.

Run it yourself: python scripts/locomo_canonical.py --conv-indices all --out results.json (~2 minutes on a laptop), or via GitHub Actions: .github/workflows/locomo.yml — 10 shards × 1 conversation, contamination-guarded aggregation, results auto-committed to benchmarks/results/locomo/.

What the remaining ~7% is (and isn't): of the ~100 remaining comparable failures, the majority are inference answers an LLM would synthesize ("What do Melanie's kids like?" → "dinosaurs, nature" — stated across scattered chunks with no shared vocabulary) and lexical variants ("names" vs "name's"). That's the honest μ=0 floor: no LLM to paraphrase-match, no fabrication. Run-to-run variance is ±0.2pp (92.94–93.35% observed; the clock pin removed the wall-clock dependency, what remains is tie-breaking on randomly-generated fact ids in the structured tier — the verbatim tier is fully deterministic, and the canonical number is the committed, sha-stamped 10-shard GitHub Actions aggregate).

How cortexm compares (search-momentum table, honest numbers)

VoiceMem (xzf-thu/VoiceMem, Aug 2026) popularized the side-by-side memory-system comparison. We borrowed the format — every competitor number below is quoted from their README/tech report, our numbers are measured, and the benchmarks are different, so rows are labeled, not conflated:

	cortexm v0.6.6	VoiceMem v0.0.1	Mem0
LongMemEval-S	500-Q full corpus: 100% (500/500) (μ=0, deterministic judge)	—	—
LoCoMo (same corpus as VoiceMem)	92.94% (1,342/1,444) — full 10-conversation corpus, single/multi/temporal, μ=0 det judge, retrieval depth k=60 (measured, 10 GitHub shards, provenance-stamped; identical code measures 92.9–93.3% across runs)	91.2% (152-Q unpublished subset, gpt-4o-mini answer + judge, Top-5)	61.68% (top-200, as reported by VoiceMem)
LLM calls at ingest	0 (μ=0 deterministic extractor)	OpenAI API required for extraction	LLM extractor required
retrieval	local, deterministic	local	cloud or local
retrieval latency (p50, warmed corpus)	~50 ms on a 636-message corpus, 2-CPU VM (1.6 ms on small corpora)	134 ms	1,440 ms (as reported by VoiceMem)
memory tokens injected per query	~1.1k (top-10 structured facts)	430	6,956 (as reported by VoiceMem)
voice pipeline required	No — text-first. Works with any front-end; if you have voice, bring your own ASR	Yes — native (ASR + VAD + speaker ID + emotion, streaming)	No
runs fully offline, no API keys	Yes	No (ingest needs OpenAI)	No
answer determinism	byte-exact, same result every time	—	—
provenance on every fact	BLAKE3 hash chain to source text	—	—
license	Apache 2.0	Apache 2.0	Apache 2.0

Why no voice? VoiceMem's pitch is memory for voice agents — it owns the ASR, voiceprint, scene, and emotion stack. cortexm's pitch is memory as a substrate: it's voice-agnostic and modality-agnostic by design. You don't need to route your users' audio through a memory system to get long-term recall — paste the transcript (or the ASR of your choice) and the trace/VSA/verbatim tiers do the remembering. If you're building a real-time voice agent and want memory co-located with the VAD loop, VoiceMem is the specialized tool; if you want deterministic, auditable memory under any front-end — text today, voice tomorrow, whatever comes next — that's this.

Known boundaries (the short list)

Full detail: docs/FAILURE_MODES.md — every failure tied to a public benchmark question.

1. The extractor is a 61-pattern lookup, not a language model. Phrasings outside the pattern library are silently dropped at ingest (e.g. "Anna has a cat named Whiskers") — they remain retrievable via verbatim/BM25 chunk recall, but never become structured facts. This is the price of μ=0: no generativity, no fabrication, no drift.
2. ZK proofs are trusted-prover attestations. The v0.6.4 backend (Pedersen + Sigma protocols on secp256k1) is sound at the commitment layer — challenges are bound to announcements, both OR-proof branches verify, H has no known discrete log, thresholds are enforced — but the linkage between committed values and store rows is established at prove-time by the prover. Verify the integration layer before trusting it against a malicious host.
3. Set membership reveals the leaf index. The value stays hidden (random-blinding Pedersen + equality proof); the position in the set does not. Position-hiding needs a ZK-friendly Merkle construction — documented future work.
4. No cross-user inference, ever. Every fact is scoped by user_id; the scope sandbox turns empty scopes into empty results (not unrestricted fallbacks). This is a feature, and it also means no "insight across users" stories.
5. Compression tiers are documented, not default. int8/binary quantization trade recall for space (see docs/COMPRESSION.md); the default build keeps full-precision embeddings because the benchmark headroom doesn't justify the loss yet.
6. Judge coverage is rule-based. The deterministic judge answers via strategy dispatch (bool/list/nugget/sum_or_diff/percentage/numeric_agg/holiday/paren). Questions outside those strategies score 0 even when retrieval succeeded — the failure is honest, the number is real.

When to use cortexm vs Mem0 / Zep / Chroma

• Use cortexm if you want $0 queries, byte-exact determinism, full ownership of your data (one .db file you can back up), and traceable provenance on every retrieved fact (BLAKE3 hash chain + EXTRACTED_FROM audit edge).
• Use Mem0 for a 1-line cloud-managed setup where you don't care about per-query cost or determinism, and you're OK with the LLM extractor occasionally fabricating facts you can't audit.
• Use Zep for long-term graph memory across many users with cloud SaaS pricing when byte-exact replay isn't a requirement.
• Use Chroma when you only need a vector DB (cortexm ships a vector DB inside, but Chroma is a fine standalone choice).

Drop-in plugins (already shipped)

• Mem0-compatible surface: from cortexm import Memory — drop-in for from mem0 import Memory
• LangChain: plugins/langchain → context-m-langchain on PyPI
• LlamaIndex: plugins/llamaindex → postprocessor
• OpenAI Agents SDK: plugins/openai_agents
• Claude Code: plugins/context-m-claude — session lifecycle hooks
• MCP server: cortexm serve (stdio JSON-RPC, zero extra dependencies)
• REST server: cortexm serve-rest — OpenAPI 3.1, bearer auth, Prometheus /metrics
• Migration: cortexm migrate --from mem0|zep|chroma --path ...

---

Documentation

The README is intentionally short. Everything else lives in docs/:

Doc	What's in it
docs/ARCHITECTURE.md	Layer 1 Symbolic Trace + Layer 2 VSA Palace + μ=0 Bridge in detail
docs/BENCHMARKS.md	Full Tier 1-4 results: OOD, in-distribution, real-GitHub, canonical LongMemEval
docs/METHODOLOGY.md	How every headline number was measured + honest scope
docs/FAILURE_MODES.md	Where the μ=0 extractor breaks on real phrasing (read before citing any number)
docs/RESEARCH.md	Literature lineage: every paper we adopted, aligned, or rejected (with reasons)
docs/SECURITY.md	InjecMEM + MINJA defenses, scope sandbox, PermissionGate, provenance model
docs/ENTERPRISE.md	PII firewall, encryption at rest, RBAC, audit, GDPR, backup/DR, REST API
docs/DEPLOYMENT.md	SDK / MCP / REST / Docker / K8s / Helm runbooks
docs/COMPRESSION.md	Storage tiers (int8 / binary / rabitq / pq) + measured trade-offs
docs/ROADMAP.md	Phase status vs the strategic plan
docs/GOVERNANCE.md	Foundation governance + licensing commitments
docs/PLAYBOOK_v2.md	Migration playbook from Mem0 / Zep / Chroma

Examples & tests

• examples/ — runnable scripts, offline, no API keys (01_quickstart → 20_agent_session)
• tests/ — 741 tests: fabric + enterprise + PPR + concurrency + sandbox + enrichment + WAL crash-recovery + migration + CRDT federation + Rust parity + ZK soundness/forgery + public-API smoke
• cortexm/experimental/ — deterministic research borrows (graph recall, coherence) — μ=0 or it doesn't ship
• leaderboard/ — self-hosted benchmark site (rebuild: python leaderboard/build.py; open leaderboard/index.html)
• AGENTS.md — how AI coding agents should interact with this repo (2026 standard)
• CONTRIBUTING.md — contribution guide

License

Apache 2.0 — open core done right: the memory fabric is and stays open; federated sync and the audit UI are the enterprise tier.