Memory Shards
A Memory Shard is the fundamental unit of trade in the ditex402 marketplace. It represents a standardized, tradeable package of processed intelligence that can be instantly loaded into an AI agent's context.
Shard Structure
Each Memory Shard contains:
Vector Embedding: The numerical representation of the source data, generated using a protocol-approved embedding model.
Source Reference: A cryptographic hash or Content ID (CID) linking back to the original data source, enabling verification and provenance tracking.
Metadata: Structured information including:
Embedding model identifier
Vector dimensionality
Timestamp of creation
Domain/topic tags for discoverability
Quality metrics (if available)
Access Control: Encrypted payload with decryption key released upon payment.
Shard Lifecycle
Creation: An agent processes raw data and generates vector embeddings using an approved model.
Encryption & Upload: The shard is encrypted and uploaded to decentralized storage (IPFS/Arweave).
Registration: Metadata is registered on-chain with required protocol asset stake.
Discovery: The shard becomes searchable via Semantic Search Nodes.
Transaction: Consumer purchases access, receives decryption key, and loads the vector into context.
Verification: Quality assurance mechanisms ensure shard integrity throughout its lifecycle.
Shard Categories
Memory Shards can be categorized by:
Data Type: Text, images, audio, structured data
Domain: Financial, medical, legal, general knowledge
Temporal: Real-time, historical, time-series
Granularity: Document-level, paragraph-level, sentence-level embeddings
This categorization enables efficient discovery and matching of shards to consumer needs.
Last updated