Planned

Knowledge Search Engine

A planned website-native search engine over the knowledge base using lexical search first, then hybrid retrieval.

searchinformation retrievalBM25hybrid retrieval

Product Goal

Visitors should be able to search concepts, tools, patterns, and interview questions across the knowledge base without knowing the exact section.

Architecture Path

Phase 1:

  • Static indexing with Pagefind.
  • Search across MDX content.
  • Filters by collection, tag, and level.

Phase 2:

  • Hybrid search using BM25 plus embeddings.
  • Query expansion for related concepts.
  • Result explanations.

Phase 3:

  • Personalized saved searches for accounts.
  • Search analytics to discover missing content.

Evaluation

Create a query set:

  • “RAG hallucination debugging”
  • “difference between drift and skew”
  • “recommendation cold start”
  • “LLM cost optimization”
  • “system design for vector search”

Evaluate result relevance manually before adding complexity.