02 / Modules

One cognitive system. Independently valuable modules.

Discover the separately testable and packageable Mentaview cognitive, confidentiality, assurance, retrieval, research, memory, ingestion and edge modules.

M4− is an internal laboratory assessment, not a production, customer, independent-assurance or certification claim.

Objective maturity review / revision 1

Every module now meets the same disclosed M4− laboratory boundary.

An implemented module exercised through a supported integration path, with repeatable local test, evaluation, verification and validation evidence, explicit failure tests, and published limitations.

11/11modules assessed at M4− · CLAIM-0042

511/511targeted counter-review tests passed · 2026-09-06

Internal labscope — no external qualification

The scale is Mentaview-specific and deliberately marked with a minus. It uses the European TRL 4 laboratory-validation boundary and NIST TEVV and software-evidence principles as external anchors, while keeping every stronger authority flag false. 511/511 targeted Rust tests passed for the four modules that were already M4-; the seven newly promoted modules remain bound to their module-specific content-addressed evidence registries. Open the complete machine-readable assessment.

01Controlled pilot

M4 internal lab

Plan the smallest sufficient path

Cognitive Engine

Interprets intent, compiles a bounded execution plan, applies budgets and selects direct inference, memory, retrieval, research, tools or explicit abstention.

Better quality-per-call, lower avoidable latency and an auditable explanation for every escalation.Open module brief

02Implemented

M4 internal lab

Know what a complete answer must contain

Question Contract Compiler

Extracts provider-neutral facets, relationships, value shapes and cardinality before evidence collection or answer publication.

Reduces false completeness and lets downstream modules repair the missing facet instead of repeating the whole task.Open module brief

03Implemented

M4 internal lab

Connect every published claim to admissible evidence

Answer Assurance

Builds a canonical evidence ledger, validates scoped references, applies per-source lexical and structural claim checks, measures requested-facet coverage and exposes precise repair gaps.

Higher citation fidelity, fewer orphan assertions and more useful abstention when available evidence cannot support the answer.Open module brief

04Implemented

M4 internal lab

Reason over aliases, not exposed identities

Confidentiality Boundary

Applies deployment-aware pseudonymization before generative inference, preserves relationships with typed aliases, and restores values only after integrity and leak checks.

Reduces avoidable disclosure of company names, people, contacts, credentials, accounts, identifiers and configured terms while preserving symbolic reasoning and exact reconstruction inside the active scope.Open module brief

05Controlled pilot

M4 internal lab

Retrieve structure, not just similar text

Document Retrieval

Consumes canonical structured evidence and combines lexical, optional semantic and conservative graph routes with bounded evidence packing, citation provenance and explicit diagnostics.

Improves coverage on large or multi-document tasks while allowing direct full-context inference to win when it is simpler and better.Open module brief

06Canary

M4 internal lab

Production-qualified fresh evidence under an explicit egress budget

External Research

Plans bounded discovery, validates targets, recovers canonical public documentation, performs safe source collection, preserves provenance and queues accepted evidence for retrieval.

Adds current public evidence without turning unrestricted browsing or a specific search vendor into a core dependency.Open module brief

07Controlled pilot

M4 internal lab

Continuity without hidden accumulation

Governed Memory

Separates recent conversation state from reviewer-approved durable facts, with scoped checkpoints, opaque source anchors, correction, retention and explicit deletion.

Creates durable continuity without silently mixing conversations, tenants, raw documents, search indexes or inferred preferences.Open module brief

08Controlled pilot

M4 internal lab

Preserve meaning across documents and media

Multimodal Ingestion

Authenticates broad source formats, discovers bounded local inputs, normalizes supported sessions, rejects exact duplicates and preserves provenance across text, image, audio and video.

Lets operators expand format coverage without surrendering provenance, process control or the right to replace — or refuse — every external converter and model.Open module brief

09Controlled pilot

M4 internal lab

Models are capabilities — never hidden dependencies

Inference Gateway

Registers explicit embedded, local, private, dedicated, public BYOK or mixed endpoints with capability, privacy, residency, health and cost metadata.

Prevents protocol names and automatic fallback from silently changing data exposure, provider class or operating cost.Open module brief

10Canary

M4 internal lab

Useful cognition when the network disappears

Edge & OEM Runtime

Composes local state, ingestion, retrieval, answer assurance, egress policy and optional separately installed inference behind provider-neutral Rust contracts.

Creates a path to appliances, desktop, embedded and sovereign environments without forcing Docker, PostgreSQL or a vendor account into the core profile.Open module brief

11Implemented

M4 internal lab

Promotion is earned, not announced

Evaluation & Release Kit

Defines frozen holdouts, replay artifacts, acceptance gates, evidence manifests, release checks and capability-specific promotion criteria.

Turns quality, latency, cost, isolation, abstention and failure behavior into release decisions instead of marketing intuition.Open module brief

Packaging rule

Buy the whole engine — or the capability that creates the value.

Mentaview's boundaries are designed to support SaaS, on-premises binaries, SDK licensing, OEM integration and selected module licensing without forcing a single infrastructure topology.

Complete cognitive runtime

Selected modules + public contracts

OEM composition + integrator policy

A deliberate next step

Start with the module that owns your bottleneck.

A pilot should prove one specific value dimension before expanding the system surface.

Start a conversation See the evidence