AnythingLLM
Let me recommend this tool — paired with ollama it can run local large models, a personal AI assistant.
Set up an LLM runtime environment with LM Studio or ollama, and then AnythingLLM can help with local document processing, vectorization, RAG, and real-time internet search before letting the model produce output.
For a while last year I ran Xunzi on my PC with it, for studying and answering questions about traditional Chinese medicine.
As for what’s so great about it — honestly, it isn’t especially powerful. When I fed my diary into AnythingLLM and tried to extract information from it, the results weren’t very satisfactory. I stopped using it after that.
But hey, it’s made the leap from 0 to 1, and the future looks promising.
LM Studio
When I tried it out last year I wasn’t using ollama — I was using LM Studio, because this thing supports MLX and can really put a Mac’s performance to work.
https://lmstudio.ai
DeepSeek
A model that stunned the world, delivering Pinduoduo-level value for money in the LLM world.
Last year when I was tinkering around I was still using a local Qwen model; this year I can use DeepSeek.
Once I have some free time, I can get back to tinkering.
CodeGPT
Since I use WebStorm and Rider, I can install the CodeGPT plugin and then use a local deployment — in theory, anyway.
MarsCode AI
I’ve already cancelled my GitHub copilot subscription, but since 2024 a lot of free domestic code assistants have come out, and MarsCode feels good enough now. There’s no need to build your own at all.