All projects · AI Agents & RAG
KB Retrieval — OpenAI + Pinecone
A semantic retrieval pipeline testing RAG query performance: vectorizes user questions with OpenAI text-embedding-3-small, performs top-K=3 cosine…
A semantic retrieval pipeline testing RAG query performance: vectorizes user questions with OpenAI text-embedding-3-small, performs top-K=3 cosine similarity search in Pinecone, and synthesizes score-ranked context blocks for downstream internal agents.
- Mirrors the 1536-dimensional latent embedding space of the ingestion pipeline using OpenAI text-embedding-3-small
- Pinecone top-K=3 similarity search scoped to the 'ryobi' namespace with includeMetadata
- Post-processor computes similarity ranks, scores, and formats a clean markdown context block for LLM prompt injection
- Serves as the foundational retrieval engine for the Puzzle #75 internal routing agent
Built with: n8n Manual Trigger, Set node (user question), OpenAI Embeddings API (text-embedding-3-small, 1536-dim), Pinecone Vector DB (/query via HTTP Request), JavaScript Code Node (context extractor & score ranker)