GlassOps Knowledge Package

This package manages the Retrieval-Augmented Generation (RAG) pipeline for GlassOps. It handles document discovery, embedding generation, vector indexing, and drift detection.

Structure

• ingestion/: Fetches and chunks markdown files from the monorepo.
• embeddings/: Generates vector embeddings using Google Gemini (via google-generativeai).
• drift/: Detects semantic drift in documentation over time.
• rag/: Provides a query engine for retrieving context-aware answers.

Usage

1. Setup

Ensure you have a .env file in the project root with your API key:

GOOGLE_API_KEY=your_key_here

2. Run the Pipeline

To scan docs, generate embeddings, and update the index:

# Using npm script (easiest)
npm run knowledge:pipeline

3. Query the Knowledge Base

To avoid argument parsing issues with npm (especially on Windows), use the Python executable directly:

# Direct Python command (Recommended)
packages\knowledge\venv\Scripts\python.exe packages/knowledge/main.py --query "What is the update policy for ADRs?"

# NPM alternative (may require extra escaping on Windows)
npm run knowledge:pipeline -- --query "What is the update policy for ADRs?"

4. Force Re-indexing

To force a full re-index:

npm run knowledge:pipeline -- --index