Skip to content

Reified RSpaces [6/11]: In-memory VectorDB implementation#333

Open
dylon wants to merge 4 commits intofeature/reified-rspaces-05-vectordbfrom
feature/reified-rspaces-06-vectordb-inmemory
Open

Reified RSpaces [6/11]: In-memory VectorDB implementation#333
dylon wants to merge 4 commits intofeature/reified-rspaces-05-vectordbfrom
feature/reified-rspaces-06-vectordb-inmemory

Conversation

@dylon
Copy link
Copy Markdown
Collaborator

@dylon dylon commented Jan 16, 2026

Summary

  • In-memory VectorDB backend implementation
  • Similarity handlers (cosine, euclidean, manhattan, dot product, etc.)
  • Distance metric implementations
  • Defragmentation for deletions (ndarray does not support efficient masked operations so the data must be defragmented)
  • Compound metric handling for combined similarity queries

This is PR 6 of 11 in the Reified RSpaces implementation.

Depends on: Reified RSpaces [5/11]: VectorDB base support and registry interface

This PR adds the complete in-memory VectorDB backend:

- RholangVectorBackend: Primary backend for Rholang integration
- Implements VectorBackendDyn trait

- in_memory.rs: Core in-memory vector index with brute-force search
- tombstone.rs: Tombstone management for soft deletions
- traits.rs: VectorBackend trait definitions

- config.rs: VectorDB configuration options
- default.rs: Default VectorDB implementation
- traits.rs: VectorDBDyn trait for dynamic dispatch

- similarity.rs: Similarity computation implementations
  - Cosine, Euclidean, Manhattan, Hamming, Jaccard, DotProduct
- ranking.rs: Result ranking and scoring strategies
- registry.rs: Handler registration and lookup
- traits.rs: Handler trait definitions
- types.rs: Query and response types

- operations.rs: SIMD-optimized vector math operations
  - Dot product, normalization, distance calculations

- error.rs: VectorDB-specific error types
- utils/binary.rs: Binary vector utilities

This is PR 6 of 11 for the Reified RSpaces implementation.
@dylon dylon marked this pull request as ready for review January 16, 2026 16:35
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant