Digest Endpoint Specification

Part of the WebMEM Protocol
Last Updated: 2025-07-28

Overview

The Digest Endpoint Exposure Layer defines how SDP-compliant fragments and full Semantic Digests are exposed for machine retrieval.

Each digest is published at a canonical, content-specific URI, and supports multi-format output using HTTP Accept: headers or query string negotiation. Endpoints are stable, cacheable, and decoupled from human-facing URLs.

URI Pattern

Primary pattern:

/semantic/[format]/[entity-id]

Examples:

/semantic/json/term-mooptotal
/semantic/ttl/plan-H5521-290-0
/semantic/prov/faq-q001

Optional content negotiation shortcut:

/semantic/[entity-id]?format=ttl

Discovery URI:

/semantic/[entity-id]/formats

Returns a list of available MIME types and representations.

Supported Formats

MIME Type	Format	Purpose
`application/ld+json`	JSON-LD	LLM pipelines, search/structured data
`text/turtle`	TTL	RDF triple stores, semantic web agents
`application/provenance+xml`	W3C PROV	Trust scoring + provenance verification
`text/markdown`	Markdown	Human-readable copy/paste, documentation
`application/xml`	XML	Enterprise and legacy systems
`text/csv`	CSV	Tabular data, spreadsheet tools

Format Negotiation

Retrieval agents can request a specific digest format via:

HTTP `Accept:` header

GET /semantic/plan-H1234-001-0
Accept: application/ld+json

Query parameter

GET /semantic/plan-H1234-001-0?format=ttl

Static path

GET /semantic/ttl/plan-H1234-001-0

Digest Discovery

Every digest endpoint may optionally expose a format manifest:

GET /semantic/plan-H1234-001-0/formats

Returns:

{
  "@id": "plan-H1234-001-0",
  "available_formats": [
    "ttl",
    "jsonld",
    "markdown",
    "xml",
    "csv",
    "prov"
  ]
}

Endpoint Rules

Canonical digest endpoints must not be coupled to page rendering paths (e.g., /medicare-advantage/az/maricopa/)
Each digest URI must be stable and versionable
Every digest must resolve to at least one format (JSON-LD recommended)
Format-specific paths (e.g. /semantic/json/) should be cacheable

Best Practices

Expose all key entities as endpoints: terms, FAQs, plans, citations, stats
Map digest URIs to human-readable fragment IDs (e.g., term-mooptotal)
Include a formats manifest endpoint for validation/discovery
Register endpoints in .well-known or GitHub for AI memory alignment