openai/chatgpt-retrieval-plugin ? reverse-engineered prompt

Build me a local version of the ChatGPT Retrieval Plugin so I can point it at my own documents and ask questions in plain English.

I want a simple backend that can ingest documents, break them into chunks, store embeddings in a vector database, and let ChatGPT or a custom GPT search them through an API. Make it easy to run locally, with clear setup instructions, environment variables, and a working example using the included local server. If there are choices for the storage backend or embeddings, wire up one sensible default and keep the rest configurable.

Please also make sure the plugin manifest and OpenAPI schema are in place, and that the main endpoints for adding, deleting, and searching documents work end to end. If anything depends on current library or OpenAI docs, look that up online and use the latest recommended approach. After you finish, I want to be able to start it, load a few sample docs, and test retrieval with a natural language query.