langchain-ai/langchain-pinecone — reverse-engineered prompt

Build me a small Python package that makes Pinecone work smoothly with LangChain for retrieval style apps. I want something that lets a developer connect to Pinecone, add documents with embeddings, update or delete them, and run similarity search or use it as a retriever inside a LangChain flow without a lot of setup pain.

Please make it feel like a real reusable integration, not just a demo script. Include clear examples that show the basic happy path, like loading some text, indexing it, then querying it back. Good defaults, helpful errors, and simple configuration for API keys would be great.

Also add a straightforward README so someone can install it, understand what it does, and get a first example running quickly. If there are current LangChain or Pinecone docs that matter, look them up and follow the modern approach so the package feels up to date.