Skip to content

feat(rag): support Redis vector search as a VDBStore backend #2170

Description

@DavdGao

Discussed in #2164

Originally posted by izualzhy July 23, 2026

Feature Request

It would be great if AgentScope could support Redis Vector Search as a VDBStore backend.

Many users already have Redis deployed in production. Supporting Redis would allow them to build small and medium-sized RAG applications without introducing an additional vector database such as Qdrant or Milvus.

Redis Vector Search provides native support for vector indexing (HNSW/FLAT), similarity search, and metadata filtering, making it a good fit for the existing KnowledgeBase + VDBStore abstraction.

From my understanding of the current design, this could fit naturally into the existing architecture as another VDBStore implementation:

                     KnowledgeBase
                           │
                           ▼
                     VectorStoreBase
                           │
          ┌────────────────┴────────────────┐
          │                │                │
     QdrantStore          ...        RedisVectorStore (new)
                                            │
                                            ├── create collection / index
                                            ├── insert documents
                                            ├── vector search
                                            ├── delete documents
                                            ├── list documents
                                            └── (metadata_filter)

This would provide a lightweight deployment option while keeping the current architecture unchanged.

If this feature aligns with the project roadmap, I'd be happy to contribute an implementation.

Metadata

Metadata

Assignees

Labels

Help WantedFurther help required.

Type

No type

Projects

Status
Todo

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions