Case Study: Character.ai achieves low-latency, high-availability search with Redis Cloud

Character.ai powers sub-250 ms search with Redis

Character.ai, a leader in AI entertainment, faced performance limitations with its previous search infrastructure built on Lancedb and Tantivy. Their system struggled with slow vector search queries, inflexible hybrid queries, and high latency during peak traffic. Additionally, their index only updated once every 24 hours, delaying new content from appearing in search results. To meet user expectations for instant, intelligent search, they turned to Redis Cloud from vendor Redis.

By implementing Redis Cloud as their high-performance data store, Character.ai unified vector, full-text, and metadata search into a single streamlined architecture. The solution delivered sub-250 ms query responses, cut index update cycles in half to roughly 12 hours, and enabled the system to handle about 650k writes per second during ingestion without impacting performance. This resulted in a 200% increase in search impressions, a 160% increase in search clicks, and a 120% increase in search conversions for new characters. Redis provided the low-latency, high-availability search needed to scale with millions of simultaneous conversations.

Character.ai