gpt-oss:20b if you can run it, it's pretty good You would use a RAG for augmenting the local model, I'll experiment with that soon

Pls share what you find out Would love to build something similar

“I help people recover lost Bitcoin using ethical hacking techniques and then professionally trade those assets to help rebuild their financial position.”

Depends on ur pc setup and what it can handle. Afaik you can run any model w ollama. I'd suggest try a lightweight one like Phi and go from there. I cant help w the data part tho, that depends on what u want to achieve.

Depends alot on the machine specs, VRAM, RAM, CPU, GPU etc. You can check model VRAM usage here: https://apxml.com/tools/vram-calculator Also people mostly run already trained models locally, since training a model from scratch requires far too much resources. You can however use tools like OpenWeb...

I don't know how experienced you are, but training a model with additional data is gonna cost you time and money. Maybe RAG System is what you are looking for. When you compare a llm whit a library, a rag system is a bookshelf with your own documents in front of the library. After a prompt, the syst...