The foundation

Technology & research

The foundations that make our solutions fast, sovereign, and verifiable — the packages, the ML engines, and the storage behind everything.

SDK · @msm-core

@msm-core — the engine layer

Thirteen pure, injectable packages. The same agent loop runs across OpenAI, Anthropic, Gemini, and local Ollama — with guards, a cost cap, and verify-and-adapt.

@msm-core/mini0.5.2

Brain-agnostic agent execution loop — guards, cost cap, force-finalize.

@msm-core/context0.3.0

Tiered RAG context assembly (zero runtime deps).

@msm-core/validate0.3.2

Rules → embedding → LLM-judge output gate; fails safe to review.

@msm-core/jobs0.8.0

CAS + idempotent durable job/mission engine, cron, HITL.

@msm-core/planner0.2.1

Verify-and-adapt planner — side-effect-aware retry or escalate.

@msm-core/learning0.1.0

Turns run outcomes into reusable lessons.

@msm-core/memory0.3.0

Five-tier memory: customer / episodic / semantic / procedural / reflection.

@msm-core/safety0.1.1

PII redaction (Luhn-checked) + prompt-injection + groundedness.

@msm-core/gate0.3.0

Human-in-the-loop approval gate + NL auto-approve.

@msm-core/rag0.2.0

Chunk · embed · search over injectable ports.

@msm-core/parse0.2.0

File → text: cloud-free parsers + born-digital PDF.

@msm-core/docx0.3.0

Real Word XML, RTL/Arabic-aware, letterhead.

@msm-core/finance0.2.0

Deterministic DCF: NPV / IRR / payback / sensitivity.

Docker · msmcore

Compute & ML workers

Heavy work ships as versioned Docker images — “packages, but for services.” Every prediction is a labelled opinion with confidence, never a hallucinated fact.

ml-tabularML / twin

train / predict / forecast / anomaly (XGBoost, LightGBM)

ml-optimizationML / twin

scheduling & allocation (OR-Tools CP-SAT)

ml-simulationML / twin

discrete-event what-if (SimPy)

ml-twinML / twin

predictive-maintenance data twin

ocr

Arabic-capable OCR (PaddleOCR)

rerank

cross-encoder reranking (ONNX, CPU)

ingest

bulk corpus: chunk → redact → embed

render

study → PPTX + XLSX

drawings

DXF / DWG CAD geometry

whatsapp-gateway

Baileys multi-session gateway (Node)

Storage · monlite

monlite — the whole backend in one file

Vectors, full-text, key-value, queue, and cron in a single crash-tested SQLite file. The reason the stack runs anywhere with no infrastructure — and the same code scales up to Postgres.

~214k/sdoc inserts
~13msvector search @100k×768
1 filevectors+FTS+KV+queue+cron
WALcrash-tested durability
@monlite/core2.11.0

Zero-dep embedded document DB (API-frozen 2.x).

@monlite/vector0.6.3

sqlite-vec / pgvector + hybrid RAG.

@monlite/fts0.7.1

FTS5 / tsvector full-text search.

@monlite/kv0.6.0

Redis-like cache, locks, sorted sets, pub/sub.

@monlite/queue0.7.0

Durable queue (SKIP LOCKED), retries/backoff.

@monlite/cron0.3.3

Persisted 5-field scheduler.

@monlite/realtime0.3.0

SSE live queries with row-level deltas.

@monlite/postgres0.2.0

Same API on Postgres / JSONB.

Documents · IntentText

IntentText — documents you can verify

A format where the file itself is the data: readable, queryable, and tamper-evident. Seal, sign, and verify offline — the trust layer a .docx cannot provide.

@dotit/core3.0.0

Parser, renderers, query, seal/sign/verify — zero deps.

@dotit/editor3.0.0

Embeddable WYSIWYG React editor (TipTap).

@dotit/pdf3.0.0

Server-side PDF + PDF/A.

@dotit/sign3.0.0

Ed25519 signatures.

@dotit/pades1.0.1

PAdES ECDSA P-256 + X.509 (Adobe-recognized).

@dotit/mcp3.0.0

MCP server: parse / query / render for agents.

Research

Research track

Beyond the platform, a model-agnostic Arabic pipeline: a semantic tokenizer (CST), a hyperdimensional intent gate (nemo), and an orchestration layer (msm) — sub-millisecond, no GPU, full EN + AR parity.

CST

Contextual Semantic Tokenizer — reversible, multilingual, Arabic root × pattern algebra.

~35–46% smaller than BPE

nemo

Hyperdimensional intent gate — 10,000-dim vectors, nearest-prototype, self-learning.

sub-ms · no GPU · EN+AR

msm

Model-agnostic pipeline — translate → brain → validate, domains declared in YAML.

swap a model in one line

Let's find your first AI win

Tell us your domain and your data posture. We'll show you a working agent — not a slide deck.