Add Redis Stack (RediSearch) as Alternative Vector Store for Code Indexing

[tupe12334]

Feature Description

Add Redis Stack with RediSearch as a user-selectable alternative to Qdrant for vector storage in code indexing. This provides users with more flexibility in choosing their vector database backend.

Motivation

• Existing Infrastructure: Many users already have Redis Stack deployed in their infrastructure
• Simplicity: Redis Stack is easier to set up than Qdrant for some users
• Performance: RediSearch's HNSW-based vector search provides excellent performance
• Flexibility: Users should be able to choose the vector store that best fits their needs

Proposed Solution

User Interface

• Add a dropdown selector in the Code Index settings to choose between:

  • Qdrant (default, for backward compatibility)
  • Redis Stack

• When Redis is selected, show configuration fields:

  • Redis URL (default: redis://localhost:6379)
  • Redis Password (optional)
  • Redis Database number (default: 0)

Technical Implementation

RediSearch Index Schema:

• Vector field using HNSW algorithm with COSINE distance metric
• TAG fields for filePath and path segments (for directory filtering)
• TEXT field for code chunks
• NUMERIC fields for line numbers

Key Features:

• Same collection naming pattern as Qdrant: ws-{SHA256(workspacePath).substring(0, 16)}
• Path segment indexing for efficient directory-scoped searches
• Automatic index recreation on vector dimension mismatch
• Metadata markers for tracking indexing completion status

Configuration

Vector Store Provider: [Qdrant ▼] or [Redis Stack ▼]

# When Redis Stack is selected:
Redis URL: redis://localhost:6379
Redis Password: ••••••••
Redis Database: 0

Backward Compatibility

• Default vector store provider remains Qdrant
• Existing installations continue to work without changes
• Switching providers requires re-indexing (no automatic data migration)

Dependencies

• ioredis - Redis client library for Node.js

[tupe12334]

🚀 Implementation Available

I've created a pull request implementing this feature:

PR #10184: feat: add Redis Stack (RediSearch) as alternative vector store for code indexing

What's Included

• ✅ RedisVectorStore class implementing the IVectorStore interface
• ✅ RediSearch HNSW-based vector similarity search with COSINE distance
• ✅ Vector store provider dropdown in Code Index settings (Qdrant/Redis Stack)
• ✅ Redis configuration fields (URL, password, database number)
• ✅ Path segment indexing for directory-scoped searches
• ✅ Automatic index recreation on vector dimension mismatch
• ✅ 41 unit tests for RedisVectorStore
• ✅ Translations for all 18 supported languages
• ✅ Backward compatible - defaults to Qdrant for existing installations

Feel free to review the PR and provide feedback!