[–] 8 points 6 months ago (3 children)

Qwen coder model from Huggingface, following the instructions there to run it in llama.cpp. Once that’s up: OpenCode and use the custom OpenAI API to connect it.

You’ll get far better results than trying to use other local options out of the box.

There may be better models potentially but I’ve found Qwen 2.5 etc to be pretty fantastic overall, and definitely a fine option beside Claude/ChatGPT/Gemini. I’ve tested the lot and it’s usually far more down to instruction and AGENTS.md instructions/layout than it is down to just the model.

  • source
  • [–] 130 points 6 months ago (5 children)

    I post there every 6-12 months in the hope of receiving some help or intelligent feedback, but usually just have my question locked or removed. The platform is an utter joke and has been for years. AI was not entirely the reason for its downfall imo.

  • source
  • [–] 5 points 7 months ago

    Success story here. 6+ years running pihole on proxmox as my primary DNS for everything on my network. It’s never missed a beat, never crashed. I update infrequently. It’s just good software.

  • source
  • [–] 2 points 8 months ago (1 child)

    Looks fantastic! I’m going to install it on my cluster this coming week for sure. Didn’t see but is there any plan to support locations or routes? Having location data along with some basic plotting could make this perfect for travel journaling too, which I’d be interested in..

  • source
  • [–] 6 points 10 months ago* (last edited 10 months ago)

    The author uses their feelings a lot to describe their distaste for MCP. It doesn’t read like a very well informed article should, in my opinion.

    MCP servers are like proxies, that can adapt their represented tools and resources based upon external conditions. They’re far from being static in nature and can provide an entry point for otherwise hidden or secured functionality. Ie. some actions may be provided via an MCP and not otherwise. File resources may be present behind an MCP server and not otherwise. Tools may be relevant for a certain agent and not others, or they may become unavailable.

    That, and regular APIs often don’t expose data in a streaming capacity that LLMs benefit from. That’s why you see MCP servers serving HTTP streams or SSE.

    Static files make this inflexible for what, simplicity?

    At least we have a standard now, for this kind of thing. Static files would be a lazy half-arsed solution at beast.

  • source
  • view more: next ›