How the AI Playground Works
A deep dive into our RAG-powered kinetic email generation system.
The RAG Pipeline
- Embedding Generation — Your prompt is converted to a vector using OpenAI embedding models.
- Hybrid Search with Pinecone — Vector database finds the most semantically similar proven examples.
- Context Assembly — Retrieved examples are formatted as reference context for the AI.
- Email Generation — Claude AI generates kinetic HTML using the examples as guidance, ensuring proper structure and lightswitch fallbacks.
Why RAG Works
- Consistent quality grounded in proven examples
- Proper email HTML with table structures and MSO conditionals
- Working lightswitch fallbacks copied from real examples
- Continuous improvement as the knowledge base grows