komikndr

What it is

This is my OpenCode setup for local models, mainly using a customized llama.cpp configuration.

For Qwen 3.8, DeepSeek V4, and Glimmer, the models are already trained to support reasoning effort levels. Depending on the model, these may be exposed as low, medium, high, xhigh, or as low, high, and max.

For models that support reasoning but were not trained with explicit reasoning-effort levels, such as the Qwen 3.5 and 3.6 variants, I use a token budget to limit the amount of reasoning.

Although Qwen 3.8 has built-in reasoning-effort levels, I still apply a maximum reasoning-token cap for each effort level.

Although the llama.cpp CLI flags specify preserve_thinking and a default reasoning budget, these can still be overridden through the API, so this works fine for my setup.

OmerFarukOruc

OpenCode + Kimi For Coding (K2.5) Setup

1. Add credentials

opencode auth login
# Select "Kimi For Coding"
# Enter your API key (sk-kimi-...)

noflcl

Here is a simple example for Seafile hosted with Docker in a Compose stack.

For your database network I create it as an --internal network: docker network create --internal your-name-here

Note: Update your email address, volume locations, and password fields

services:
  seafile-db:
    image: mariadb:10.11

jwaiswa7

REFRESH TOKENS WITH DEVISE

I have been using the devise gem with rails to implement user authentication for may applications. When implementing authentication though API requests, then I turn to the devise-jwt gem. I faced an issue when implementing refresh tokens with devise, as devise does not support access tokens. I still needed the superior authentication that devise provides, but needed to manually work with fresh tokens. After looking around the web, here is how I was able to modify the code to implement refresh tokens with devise-jwt.

Take for the example a user model to sore the user email and password.

veekaybee

Anti-hype LLM reading list

Goals: Add links that are reasonable and good explanations of how stuff works. No hype and no vendor content if possible. Practical first-hand accounts of models in prod eagerly sought.

Foundational Concepts

Pre-Transformer Models

gabrc52

#!/bin/bash

# Debug
set -o xtrace

GUEST_IP=192.168.122.221
HOST_IP=192.168.122.1

# Wipe the current usbfluxd, usbmuxd, and socat:
sudo killall usbfluxd

justin808

# Using these pry gems -- copy to your Gemfile
# group :development, :test do
#   gem 'awesome_print' # pretty print ruby objects
#   gem 'pry' # Console with powerful introspection capabilities
#   # pick either:
#      # using byebug, but has issues with Zeitwerk
#      gem 'pry-byebug' # Integrates pry with byebug
#
#      # using default ruby debuggger
#      gem 'pry-stack_explorer'

searls

class Whereable
  def initialize(where:, model: Item, ranking_conditions: [], valid: true, data_source: nil)
    @model = model
    @where = where
    @data_source = data_source
    @ranking_conditions = ranking_conditions
    @valid = valid
  end

  def valid?

munificent

#include <time.h> //  Robert Nystrom
#include <stdio.h> // @munificentbob
#include <stdlib.h> //     for Ginny
#define  r return    //    2008-2019
#define  l(a, b, c, d) for (i y=a;y\
<b; y++) for (int x = c; x < d; x++)
typedef int i;const i H=40;const i W
=80;i m[40][80];i g(i x){r rand()%x;
}void cave(i s){i w=g(10)+5;i h=g(6)
+3;i t=g(W-w-2)+1;i u=g(H-h-2)+1;l(u

simonbroggi

using System;
using System.Collections;
using System.Collections.Generic;
using UnityEngine;
using UnityEditor;

public class BuildAutomation {
	public struct BuildTargetAndGroup {
		public BuildTargetGroup group;
		public BuildTarget target;