Start trial

Domain & Feature Packs

Make AthenaDB a specialist in your field

AthenaDB starts general on purpose. A pack teaches it a domain — the right structure, how to read your documents, and the workflows that matter — as a signed add-on that runs in a strict sandbox, so it only ever touches what it's allowed to. Legal, Medical, and Research are ready today.

Available nowIn progressPlanned

How a pack works

  1. 1

    Manifest

    A pack declares its object types, relationships, embedding namespaces, and the type-tags it is authorized to touch — plus the core version it needs. The core engine stays domain-agnostic.

  2. 2

    Sign & seal

    The manifest is cryptographically signed into a sealed package. The signing key is held in a secure vault and never ships with the pack — so you can trust that a pack is exactly what its author published.

  3. 3

    Verify & load

    At load time the runtime checks the signature and a core-version compatibility bound before allocating anything. An unsigned, tampered, or too-new pack fails closed and is never registered.

  4. 4

    Sandbox

    Every pack touches data only through a sandbox that enforces its declared scope — reads and vector searches are limited to authorized tags and owned namespaces. Out-of-scope access is denied and audited. A pack can never bypass encryption, permissions, or audit.

Today the first-party packs ship compiled into the engine and are invoked as CLI subcommands. The signed seal → load → sandbox mechanism that makes packs true add-ons is built and tested; dynamic install and pack purchase are on the roadmap.

Pack families

Domain packs

Available now

Vertical schema, ingestion, workflows, and RAG policy — Legal, Medical, Research today.

Embedding packs

Planned

Specialized embedding models and vector configurations for a domain.

Connector packs

Planned

Ingest from M365, Google Workspace, GitHub, Jira, ServiceNow, MyChart, email.

Compliance packs

Planned

Policies, audit templates, retention profiles, and export controls (HIPAA, CJIS-style, legal hold).

Deployment packs

Planned

AD/Entra, Azure, AWS, air-gapped, and field-gateway deployment profiles.

Domain packs available today

Legal

Available now

A litigation-grade case knowledge graph.

The deepest pack: a proof chain of Evidence → Fact → Element → Claim across 19 object types, with fact ledgers, a versioned case theory, discovery and privilege tracking, deadline computation, drift detection, and pre-filing checks.

Explore

Medical

Available now

Encounter-centric clinical records.

An EHR-aware pack built around the clinical encounter: 11 object types with FHIR R4, C-CDA, and MyChart/Epic ingestion, plus problem, medication, lab, vital, allergy, and immunization ledgers.

Explore

Research

Available now

Scholarly works, claims, and citations.

A research-corpus pack modeled on OpenAlex/CiTO/SciCite/SciFact: works, authors, venues, concepts, reified citation edges with intent, and SciFact-style claims — with deterministic reference parsing.

Explore

On the roadmap

Construction

Planned

Project, submittal, and RFI modeling. Planned.

Government

Planned

Public-records, FOIA, and retention workflows. Planned.

Cybersecurity

Planned

Incident, artifact, and IOC modeling. Planned.

Building a pack

The manifest is the contract: declare object types, relationships, and namespaces; implement deterministic extractors, rules, workflows, and context templates against the pack interfaces; sign it. The runtime loads it and confines it in a sandbox. The framework is built for third-party packs; a packaged, documented public SDK and third-party trust model are on the roadmap.