Add Time-to-Live (TTL) Support to RedisVectorStore Index

[hatemalimam]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

Introduce Time-to-Live (TTL) support for vectors and their associated metadata in the RedisVectorStore index. This feature would allow developers to specify a TTL for data entries upon insertion, ensuring that outdated or irrelevant data is automatically cleaned up after a predefined time.

Motivation

Managing the lifecycle of vector data is crucial in dynamic use cases such as real-time recommendation systems, session-based personalization, and temporary data storage. Without TTL, developers must implement custom logic to periodically delete expired data, which increases complexity and overhead.

By providing TTL support directly in the RedisVectorStore index, users can:

Reduce the manual effort required to manage data expiration.
Improve the performance of queries by automatically purging stale data.
Ensure better resource utilization by preventing the accumulation of unused data.
This feature aligns with Redis's native capabilities of supporting TTL for keys and would extend this functionality to vector-based data management.

Proposal (If applicable)

• Add an optional ttl parameter to the data insertion method in RedisVectorStore.
• The ttl parameter should accept values in seconds to indicate the time period after which the entry will expire.
• Upon setting a TTL, Redis should automatically delete the vector and its associated metadata once the TTL expires.
• Provide a utility method to update or remove the TTL for existing entries if needed.

RedisVectorStore(this.embeddings, {
      redisClient: client,
      indexName: `index_name`,
      ttl: 600 // 10 minutes in seconds
});