chroma-core/chroma ? reverse-engineered prompt

Build me an open source search database for AI apps that I can use locally right away and also run as a simple server when I need it. I want it to feel really easy from Python, and ideally JavaScript too, with a tiny API where I can create a collection, add documents with ids and metadata, update or delete them later, fetch by id, and run similarity search on natural language queries.

By default it should handle the embedding and indexing work for me, but I also want the option to pass my own embeddings. Search results should support basic filtering on metadata and simple text contains style matching inside documents. Please make persistence easy so I can start in memory for testing and then save to disk without much hassle.

Include a quick way to run it locally, a short example app, and clear docs or examples that show the happy path. If anything is unclear, look up the current docs online and wire it up in the most practical way.