xhillman/synapticdb ? reverse-engineered prompt

Build me a small Python library that gives an AI assistant a persistent memory store in one SQLite file.

I want to be able to save short notes, search them later by both keywords and meaning, and get back a confidence score so the app can ignore weak matches instead of always pretending it found something useful. It should also let me attach simple metadata, filter by that metadata, connect related memories, and record feedback when a result was useful or not. If possible, include a built in local embedding option, but also let me plug in my own embedding function.

Please make it easy to use from code with something like storing a memory, recalling relevant context for a question, and exporting results as JSON. Keep it lightweight and self contained, with everything living in the database file. If there are good default examples or a tiny demo for using it with an AI chat request, include those too. If you need to check current docs for anything, go ahead and look them up online.