[FEATURE] AI 5 : RAG Pipeline - Retrieval and Reasoning Agent

[TFT444]

Objective

Build the retrieval and reasoning layer that uses the vector store to ground AI responses in actual OpenShield rules, CIS controls and NIST controls rather than general knowledge.

What needs to be built

• Retrieval step that fetches relevant context before generation
• Prompt construction that injects retrieved context
• Agent reasoning over findings combined with retrieved knowledge
• Replaces the simple completion call from Phase 1

Acceptance criteria

• Responses grounded in OpenShield rules and compliance controls
• Works with all three providers
• Retrieval step returns relevant context for given findings

Dependencies

Depends on Issue 4 (document ingestion and vector store)

Priority

Phase 2