Part of the WebMEM Protocol
Location: /specification/sdt/
Last Updated: 2025-07-28
Semantic Data Templates
Semantic Data Templates (SDTs) are purpose-built containers for exposing structured, trust-scored memory fragments to AI agents, indexing systems, and retrieval pipelines.
At their core, SDTs answer the question:
How can we publish machine-readable facts and definitions that are independently retrievable, provenance-scored, and glossary-aligned—without disrupting the human-facing web?
Each SDT fragment encodes a precise claim or definition, scoped to a real-world entity (e.g., plan, provider, term), and grounded in transparent ProvenanceMeta. These fragments operate as atomic memory units—machine-ingestible, versioned, and independently retrievable.
Why Do SDTs Matter?
SDTs are more than a publishing format—they’re a memory protocol. They matter because they:
- Elevate Transparency
Every fact or term is scoped, sourced, timestamped, and glossary-linked. - Enable AI Reflection and Retrieval
Fragments include structured fields that allow AI agents to cite, retrieve, and reason over discrete memory units with clarity and trust. - Power Digest Publishing
SDTs form the backbone of WebMEM digests—curated, structured outputs optimized for machine access, scoring, and feedback integration. - Support Provenance and Auditability
SDTs natively embed W3C PROV metadata and support fragment-level scoring (per RFC-005). - Modularize Memory Exposure
Each SDT fragment is an atomic memory object—validatable, exportable, and independently usable by agentic systems.
From Static Pages to Structured Memory
SDTs shift the web from static content to semantic memory. They allow organizations to share facts, definitions, and logic not just for humans to read—but for machines to embed, cite, and retain.
If the web was built for visibility, SDTs are built for memory.
What is an SDT?
A Semantic Data Template (SDT) is the foundational memory fragment format of the WebMEM Protocol.
Each SDT is:
- Trust-scored
- Glossary-aligned
- Provenance-anchored
- Retrievable across formats and interfaces
There are two primary embedding formats:
- YAML-in-HTML — full fragment format inside
<template>blocks - Data-in-HTML — inline markup using
data-*attributes
SDT Specification Modules
| Section | Description |
|---|---|
| Overview | How SDTs work and why they matter |
| Fragment Classes | All supported data-sdt-class types |
| YAML-in-HTML Format | Template-based fragments with full metadata |
| Data-in-HTML Format | Inline fragments using HTML attributes |
| Attributes & Fields | Full schema reference (YAML + data-*) |
| Trust Layer Integration | ProvenanceMeta + scoring logic |
| Validation Rules | Compliance requirements + error handling |
| Export & Digest Endpoints | TTL, JSON-LD, Markdown, PROV, and more |