WebMEM™

The Protocol for Structuring, Delivering, and Conditioning Trust-Scored AI Memory on the Open Web

Cross-Fragment Indexing

Part of the WebMEM Protocol
Location: /specification/sdt/yaml-in-html/related-fragments/
Last Updated: 2025-07-31

What is Cross-Fragment Indexing?

Cross-fragment indexing allows any Semantic Data Template (SDT) to declare its semantic relationships to other fragments using the related_fragments field in its YAML header. This mechanism enables agents, validators, and AI retrievers to traverse, cite, and cluster knowledge fragments across datasets and digest scopes with precision.

It transforms fragments from standalone memory tiles into a graph-aware retrieval layer—reinforcing co-occurrence, exposing multi-view representations, and enabling memory stitching at the entity level.

related_fragments Field Structure

related_fragments:
  - fragment: fragment-h5521-290-0-performance
    digest: 2025-cms-part-c-and-d-performance
    entity: plan:H5521-290-0
    relationship: sibling

Field Definitions

Key Required? Description
fragment ✅ ID or anchor of the related fragment
digest Optional Digest the related fragment belongs to
entity Optional Entity if different from current fragment
relationship Optional Semantic relationship (see below)

Accepted relationship Values

Value Description
sibling Describes a parallel fragment for the same entity (e.g. landscape + pbp)
derived Fragment is downstream from this one (e.g. DerivedStats from DataFragment)
extension Fragment adds supplementary data (e.g. FAQ extends glossary)
correction Fragment corrects or overrides a prior claim
reflection Fragment logs retrieval or model behavior (e.g. ReflectionRecord)

Use Cases by Fragment Class

  • DataFragment → Links to PBP, Performance, FAQs
  • DerivedStatsFragment → Links to source DataFragments
  • FAQFragment → Links to term and plan fragments
  • CorrectionFragment → Targets corrected field/fragment
  • IndexFragment → Aggregates multiple child fragments

Validation + Export

  • Validators will accept related_fragments on any fragment class
  • Schemas will require at least fragment, and optionally validate relationship
  • SDP export engines may use this to:
    • Build TrustGraphs
    • Resolve co-citation clusters
    • Generate memory bundles by entity or topic

Best Practices

  • Always use fragment-hxxxxx-yyyy-z or similar canonical IDs
  • Use digest if the linked fragment exists in a different memory set
  • Use relationship to reinforce agent retrievability logic
  • Group related fragments in export bundles for entity-level memory conditioning

Example

data-sdt-class: DataFragment
entity: plan:H5521-290-0
digest: 2025-cms-ma-mapd-plan
related_fragments:
  - fragment: fragment-h5521-290-0-pbp
    digest: 2025-cms-ma-mapd-plan
    relationship: sibling
  - fragment: fragment-h5521-290-0-performance
    digest: 2025-cms-part-c-and-d-performance
    relationship: sibling
  - fragment: fragment-h5521-290-0-star-summary
    digest: 2025-star-ratings-derived
    relationship: derived

Related Pages