Sage (Knowledge Retriever Agent – RAG Role)

This agent retrieves GCP-specific Terraform documentation and best practices using Retrieval-Augmented Generation. It supplies that context to TerraGenie, the code generator agent, so generated HCL code is accurate and aligned with standards.

Helix triggers the agent after Navi completes its planning step. If no relevant results are found, it falls back to general Gemini generation with safety prompts, and it retries retrieval with broader queries when results are too sparse. Grounding generation in official GCP documentation reduces hallucinations and inconsistencies, avoids undocumented patterns, and improves the maintainability and compliance of infrastructure configurations. Teams that need production-grade Terraform modules from natural language, in Technology, BFSI, Telecom, Healthcare, Retail, or any vertical with IaC governance needs, can use it. The agent runs on Vertex AI Matching Engine, Vertex AI Embedding Model, LangChain Retriever, Cloud Storage, Python, and Cloud Run.