Previous document
Keep the content identifier of the version used in the last review.
TOKENISED ASSETSReal assets.
Real assets.
Traceable knowledge.
Connect issuer research, document references and asset updates. Keep the information behind each decision addressable and verifiable.
Observe→Recall→Record
MEMORY IN PRACTICE
Every asset has a story beyond its token.
Issuer documents, research notes and updates contain context that a token identifier cannot carry on its own. Give research agents a durable record of the information behind an asset.
EXAMPLE WORKFLOWTokenised assets
USER REQUEST
Start with the question.
An issuer has published an update. What has changed since our last review?
Start with the question.WHAT THE WORKFLOW CAN REMEMBER01 02 03
Research notes
Record the assumptions and open questions from that review.
Asset context
Link the issuer reference and document history to the asset identifier.
Example inputs and responses for this workflow.
THE PROBLEMWhen the context
When the context
doesn’t carry forward.
Documents lose their version
An updated link can point to different content from the document an earlier decision relied on.
Research starts from scratch
A new session has the asset identifier but not the issuer context or unresolved questions.
File access and memory get confused
Sensitive source files need explicit access handling, separate from the summaries and references used by an agent.
WHAT MEMORY MAKES POSSIBLEKeep what matters.
Keep what matters.
Build on what came before.
01DOCUMENT REFERENCES
Document versions you can reference
Record the content identifier of the specific document used in a review. Retrieve that content later and verify that the bytes match the reference.
Asset ID
Document CID
Review date
02ISSUER MEMORY
Issuer context across sessions
Persist research notes and source references in a consistent namespace. Keep the assumptions behind an assessment available for the next review.
Issuer updates
Research notes
Open questions
03ASSET HISTORY
An update trail that stays connected
Write new asset updates as additional records linked by tags and identifiers. Your application can compare new information with the earlier history.
Earlier version
New update
Change review
04ACCESS DESIGN
A deliberate boundary for sensitive files
Use Lighthouse’s existing encrypted file storage where appropriate for source documents. Treat encryption for the memory rail as a separate roadmap capability.
Encrypted source file
Document reference
Selected context
THE WORKFLOWFrom a new observation
From a new observation
to lasting context.
Reference the source document
Store the document and keep its CID with the asset identifier and review timestamp.
Recall the previous review
Retrieve relevant issuer notes, document references and unresolved questions.
Preserve the new context
Write the updated assessment while retaining the references behind earlier decisions.
Content integrity does not establish ownership, valuation or legal validity of an asset. Those checks belong to your application.
DESIGN YOUR INTEGRATIONA memory layer.
A memory layer.
Your application’s logic.
Connect the memory runtime to the moments your application needs to write, recall and review context.
Explore memory documentation01
Use a stable asset identifier in your record tags.
02
Retain the exact document version behind each review.
03
Separate file access decisions from the context passed to the model.
04
Keep ownership, valuation and compliance checks in the application.
COMMON QUESTIONS
Before you build.
Does a CID prove ownership of an asset?
No. A CID identifies content. It does not establish legal title, valuation, authenticity of an issuer or rights attached to a token.
Can we keep earlier document versions?
Your workflow can store each version and retain its content identifier. Keeping those references lets your application retrieve the document used in an earlier review.
Is memory encrypted in the same way as files?
The whitepaper distinguishes existing encrypted file storage from planned Kavach-backed encryption for memory. Do not assume that the two paths currently offer identical capabilities.
What belongs in an asset research memory?
Examples include issuer notes, source references, document CIDs, review timestamps, updates and open questions. Choose records that support the next review.