ifoxhz

I'm now using this approach too, and it feels better than any sorting method I've used before. The only thing I'm thinking about now is: if the LLM makes a mistake, how can I provide feedback and verify it?

That’s a lot of fun!

good idea. I also hope this.

You said it!

Since I don't have any macOS environment, contributions for macOS build capabilities are welcome.

Built upon the Pi agent, piFlow lets you import local documents to build a private knowledge base. Ask questions in natural language, and get answers with source citations—every response traces back to the specific document and paragraph. All documents, vector indexes, and chat history are stored in a local SQLite database, with zero cloud dependencies.

Supports PDF (including scanned documents with OCR), Markdown, TXT, HTML, and more. You can choose your generation backend: local GGUF, remote Ollama, or DeepSeek. The app is built with Tauri 2 + Node dual‑process architecture. continuously improving it — contributions and feedback are always welcome!

I think so.

FUSE generally has low overall performance because of an additional data transfer process between the kernel space and user space, which is less than ideal for AI training.

I completely agree with your point. I also strongly dislike this programming model. However, are there better handling mechanisms or well-established libraries for managing concurrency and synchronization in Go? Previously, when I used C, I relied heavily on libraries like libuv to handle similar issues.

"Learn to walk before you run. I firmly believe in this wisdom. First build a foundational model, then tackle boundary challenges and efficiency optimization."