AgentDB Advanced Features FAQs

Question 1

When should I use Euclidean distance vs Cosine similarity?

Accepted Answer

Cosine similarity is best for general-purpose semantic search and text embeddings, while Euclidean distance (L2) is preferred for spatial data, image embeddings, and scenarios where vector magnitude is significant.

Question 2

How does Hybrid Search improve results?

Accepted Answer

Hybrid search allows you to combine the semantic power of vector similarity with strict metadata constraints (like dates, categories, or price ranges), ensuring results are both conceptually relevant and factually accurate.

Question 3

Can I shard my AgentDB database for scaling?

Accepted Answer

Yes, this skill provides patterns for horizontal scaling by sharding databases by domain or key, allowing you to distribute data across multiple files or nodes to maintain performance as your dataset grows.

Question 4

What is the benefit of using MMR (Maximal Marginal Relevance)?

Accepted Answer

MMR helps avoid 'echo chambers' in AI retrieval by balancing result relevance with diversity, ensuring the AI receives a broad range of information rather than multiple redundant copies of the most similar result.

Question 5

What is QUIC synchronization in AgentDB?

Accepted Answer

QUIC synchronization enables sub-millisecond latency when syncing data between multiple AgentDB instances across network boundaries, featuring automatic retries, multiplexing, and built-in TLS 1.3 encryption.

AgentDB Advanced Features

AgentDB Advanced Features

Key Features

Use Cases

Key Features

Use Cases