Meaning-based search for the newsroom

A SaaS B2B business working with Cloud Combinator on AWS. The client is anonymised at their request.

The challenge had three focus areas, each about turning a growing article library into search that understands intent.

New content flows in through Amazon S3. When a valid document is uploaded to the articles bucket, an ingestion Lambda converts it into vectors using Amazon S3 Vectors and loads it into an Amazon Bedrock Knowledge Base, so the searchable index keeps itself current as articles are published. An one-time Lambda, triggered on deployment, initialises the S3 Vectors store first, since it must be created inside AWS once the resources exist.

The integration was delivered as scoped: a serverless, event-driven search capability that retrieves articles by meaning, keeps its index current automatically and connects to the frontend through a single API.

• Meaning-based retrieval — Amazon Bedrock compares a reader's query against the Knowledge Base and returns the articles whose content and metadata match the intent, not just the keywords.
• Self-updating index — A S3-triggered ingestion Lambda vectorises and loads each new article automatically, so search stays current with no manual re-indexing.
• Serverless by design — Every component runs on Lambda, S3 and managed services, scaling with demand and costing nothing when idle.
• Simple API contract — Amazon API Gateway exposes a single query-in, links-out endpoint that drops into the existing frontend.
• Handover — The search integration was delivered ready to run in the client's environment, with the ingestion and retrieval flow validated end to end.

With a Bedrock-backed search foundation in place, the client can keep expanding its library while search quality holds up, and the same pattern is ready to extend into richer retrieval and generative features as the product grows.