Skip to content
OKF-first knowledge catalog

Compendium

Connect AI agents to the documents your enterprise already has, and turn them into a living, agent-curated map of the business.

Get Started GitHub

What is Compendium?

Compendium helps enterprise architects connect AI agents to the documents their organization already has — scattered across SharePoint sites, Confluence spaces, wikis, and shared drives — and turn them into a living, agent-curated map of the business.

  • OKF-First

    Every concept is a plain markdown file with YAML frontmatter, conforming to the Open Knowledge Format — readable without Compendium, version-controlled in git, and portable with no proprietary store in between.

    OKF Format

  • Multi-Source Connectors

    Connect to SharePoint sites, Confluence spaces, markdown wikis, local file systems, and git repositories, transforming source documents into OKF concepts while preserving provenance.

    Connectors

  • Data Lineage Tracking

    Built-in data maps document field-level data flows — automatic detection and grouping by integration, source/destination tracking, transformation logic, and graph-based lineage queries.

    Data Lineage

  • AI Agent

    The Compendium system agent curates the catalog — ingesting documents, creating concepts, linking relationships, flagging stale knowledge — and answers questions grounded in the catalog with traceable sources.

    AI Agent

  • 14+ File Formats

    Ingest documents (.txt, .md, .pdf, .docx), data (.json, .xml, .csv, .xlsx), email (.eml, .msg, .ost), diagrams (.drawio, .vsdx), and architecture models (.archimate).

    Supported Formats

  • Trustable at Scale

    Provenance, trust, and lifecycle fields keep knowledge honest as more of the catalog is written by agents rather than people — every concept knows where it came from and whether it's gone stale.

    Concepts

Quick start

./build.sh
compendium init
./bin/web/Compendium.Web

.\build.ps1
compendium init
.\bin\web\Compendium.Web.exe

Or use the CLI directly:

compendium chat --bundle catalog/sample

Full installation guide

Documentation

Doc What it covers
Installation Prerequisites, build, first run
Quick Start Up and running in 5 minutes
Configuration LLM provider setup and config schema
Concepts The anatomy of an OKF concept
Ingestion Turning source documents into concepts
Data Maps Field-level data lineage across integrations
Chat Interface Querying and curating the catalog conversationally
Web UI The Blazor Server catalog browser
CLI Command-line reference
OKF Format The Open Knowledge Format bundle spec
Connectors SharePoint, Confluence, wikis, file systems
Data Lineage Data governance and impact analysis
AI Agent Catalog curation and grounded Q&A
Building Building Compendium from source
Architecture System design and component interactions
Contributing How to contribute code and docs
Supported Formats Format support matrix
API REST API reference

Architecture

graph LR
    subgraph Sources["Enterprise Sources"]
        SP[SharePoint]
        CF[Confluence]
        MW[Markdown wikis]
        OT[...]
    end

    subgraph Compendium["Compendium"]
        CN[Connectors]
        EC[Enrichment / curation]
        BN[OKF bundle<br/>git]
    end

    subgraph Consumers["Consumers"]
        AG[Compendium<br/>system agent<br/>Q&A + curation]
    end

    SP -->|ingest| CN
    CF -->|ingest| CN
    MW -->|ingest| CN
    OT -->|ingest| CN

    CN --> EC
    EC --> BN
    BN --> AG

Use Cases

  • Enterprise Architecture Documentation — Catalog systems, processes, and integrations
  • Data Governance — Track data lineage and field-level flows
  • Knowledge Management — Connect scattered documentation into a coherent graph
  • AI-Assisted Research — Enable agents to answer questions with source attribution
  • Onboarding — Help new team members understand the enterprise landscape

Status

Compendium is in early preview. See the GitHub repository for current progress and roadmap.

License

Open source software. See LICENSE for details.