Purpose of the change

Motivation

• current DeclarativeMemory does not handle multimodal content
• current DeclarativeMemory does not handle chunking
• current DeclarativeMemory produces high fan out in number of Neo4j queries
• current DeclarativeMemory operations are not very atomic
• VectorGraphStore is difficult to implement using other databases
• we have a new VectorStore interface defined previously to move toward a solution
• tangential: current DeclarativeMemory may face problems with top-k redundancy from vector search

Description

Designed to scale better than existing Neo4j implementation -- need concrete implementation to verify.

Changes:

• define new extensible data models to support multimodal content and chunking
• define new APIs to allow for more efficient and atomic operations

Approach to ingestion:

• derivatives may or may not be deduplicated/consolidated
  • vector search hits a representative, which has multiple segments linked to it
  • some filters may be applied before, some filters may be applied after -- depending on whether the filter changes semantic meaning

Approach to search:

• derivatives are no longer filterable
• for low cardinality filters, we can get all entries from DB then do a brute force vector search
• for high cardinality filters, we can do a vector DB search then post-filter
• for no filters, we can do a vector DB search

Approach to deletion:

• API designed with SQL in mind
• use reference counting, active and purging states
  • purging state necessary as a lock to allow deleting from external DB (vector DB)

Alternatives considered:

• return UUIDs of vectors to delete when vectors become orphaned: synchronous deletion does not scale well -- we will instead implement a garbage collection job -- delete will be soft until garbage collection job

Decisions to make:

• whether this will replace existing memory or just be a new, better memory
• naming

TODO:

• implement new memory (basically the same basic logic as DeclarativeMemory but should be better)
• change content to discriminated union or similar

Type of change

Will eventually be breaking or require new API.

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Refactor (does not change functionality, e.g., code style improvements, linting)
• Documentation update
• Project Maintenance (updates to build scripts, CI, etc., that do not affect the main project)
• Security (improves security without changing functionality)

How Has This Been Tested?

APIs and data models not tested.

• Unit Test
• Integration Test
• End-to-end Test
• Test Script (please provide)
• Manual verification (list step-by-step instructions)

Checklist

• I have signed the commit(s) within this pull request
• My code follows the style guidelines of this project (See STYLE_GUIDE.md)
• I have performed a self-review of my own code
• I have commented my code
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added unit tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• Any dependent changes have been merged and published in downstream modules
• I have checked my code and corrected any misspellings

Maintainer Checklist

• Confirmed all checks passed
• Contributor has signed the commit(s)
• Reviewed the code
• Run, Tested, and Verified the change(s) work as expected