Hey,

Article 5 of the AI-Native IDP series is live:

TechDocs RAG — AI That Reads Your Docs

Your AI assistant can write code. But ask it "what's our retry policy for Service Bus?" and it has no idea. The answer is in your TechDocs — page 14, section 3.2. Nobody reads it. Not because developers are lazy, but because finding the right paragraph takes longer than asking in Slack.

In this article:

• A vector store (PostgreSQL + pgvector) that indexes all your platform documentation

• An embedding pipeline that splits docs into chunks and stores them as vectors

• A search endpoint that finds relevant docs using cosine similarity

• An "Ask about this service" widget on the Backstage entity page

• Works with any embedding model — Scaleway, OpenAI, Azure AI Foundry

The AI answers from your actual docs and cites the source. No hallucination, no guessing.

Code is on GitHub: victorZKov/forge

Read it here: TechDocs RAG — AI That Reads Your Docs

Next up: The AI Governance Dashboard — who uses what, how much it costs, and who controls it.

Victor