via “retrieval-augmented generation (rag) with configurable engines”
Agent framework returning Design, Tasks, or Repo
Unique: Supports multiple RAG engines (vector search, BM25, hybrid) with pluggable configuration, enabling teams to choose the best retrieval strategy for their use case. Retrieved context is automatically injected into prompts with source attribution, enabling agents to cite sources and enabling verification of retrieved facts. RAG configuration is declarative, allowing different agents to use different knowledge bases without code changes.
vs others: More flexible than single-engine RAG systems because it supports multiple retrieval strategies and knowledge sources, enabling teams to optimize for their specific domain. Hybrid retrieval (combining vector and BM25) provides better recall than vector-only approaches, reducing the risk of missing relevant context.