What problem does it solve? Agents lose context between sessions, and building retrieval-augmented generation pipelines requires wiring together embedding models, vector indexes, and query logic. This Skill provides ready-to-use scripts and patterns for persisting agent memory and retrieving relevant context with Pinecone. ## Core Features & Use Cases - Session-Scoped Memory: Store and recall conversation memory in isolated Pinecone namespaces per session, with cross-session search available through the global namespace. - Document RAG Pipeline: Index directories of text documents into a Pinecone index with batched upserts, then query them with embedding-based similarity search. - Index Management: Create serverless indexes on demand, inspect vector counts per namespace, and clean up stale session data to control costs. - Use Case: An agent handling multi-day research sessions stores each day's findings under a session namespace, then recalls relevant prior decisions when the user asks about earlier architecture discussions. ## Quick Start Set the PINECONE_API_KEY and OPENAI_API_KEY environment variables, then ask the agent to store a memory or query the agent-memory index using the provided scripts.
