mari-components · 0.1.0.dev0

Mari documentation

Choose a system boundary

Area

What it covers

Ingest

Documents, multimodal regions, code structure, polling/streaming connectors, synchronization, parsers

Retrieve

Lexical, dense, multi-vector, approximate, contradiction, graph, adaptive, and lifecycle retrieval

Govern

Evidence, trust-gated writes, source conflicts, retention, freshness, verification

Memory

Admission, mutation, scopes, promotion, segmentation, salience, organization, consolidation

Graphs

Semantic schemas, constraints, resolution, temporal facts, propagation, communities, projections

Agents

Trajectories, procedure mining, regression gates, reviewed reuse

Platform

Portable bundles, living views, artifact lineage, stores, pipelines, task evaluation, compilation

Mari is a framework-neutral Python library of composable tools for extracting, validating, linking, retrieving, comparing, and evaluating knowledge from changing source material. It does not prescribe a knowledge graph, ontology, storage system, construction pipeline, agent framework, or model. Operations accept caller-owned values and return immutable candidates, scores, plans, reports, and traces.

Current — importable from mari_components

How to use these docs

Each feature page describes importable code in mari_components. Research-derived mechanisms link evidence next to the explanation; application-injected model calls, persistence boundaries, and failure behavior are identified explicitly.

Package namingmari-components is the distribution. Public imports use mari_components.