The initial architecture for the finlight.me real-time financial news API utilized Postgres full-text search, which was effective for early needs but faced limitations as article volume increased. Key challenges included performance degradation due to the inability to combine GIN indexes with regular B-Tree indexes, leading to slow query performance, especially with complex user queries and pagination. The decision was made to transition to OpenSearch for its optimized search capabilities while retaining Postgres as the source of truth for data integrity. Testing with OpenSearch revealed the need for production-grade resources, which led to a successful deployment that improved search response times significantly. The current architecture separates ingestion, storage, and retrieval responsibilities, with Postgres managing data storage and OpenSearch handling user-facing search queries. Regular snapshotting and thoughtful index management ensure resilience and scalability.