second-state

/

jina-embeddings-v2-base-code-GGUF

like 4

Follow

Second State 257

So I've been building a code-aware MCP server [https://github.com/Superuser666-Sigil/SigilDERG-Custom-MCP] that currently uses the MiniLM model for sentence transformers to do embeddings for my local repos. This model comes highly recommended by many sources I've spoken with as a much better solution for this, and I'll be trying it out today to see what kind of improvements I get over the MiniLM model.

That said, are there any quirks I should be aware of before using this in prod, or tasks it excels at versus stinks at, such as programming languages it does better with, how it handles large groups of ADRs, runbooks, and other documentation, etc.?