ben-vargas

  {
    "custom_models": [
      {
        "model_display_name": "Claude Opus 4.5 [Proxy]",
        "model": "claude-opus-4-5-20251101",
        "base_url": "http://localhost:8317",
        "api_key": "dummy-not-used",
        "provider": "anthropic"
      },
      {

ben-vargas

Amp Model Override Cheat Sheet

How to Override the Default Model

Option 1: amp.model.sonnet (simple toggle)

{
  "amp.model.sonnet": true
}

ben-vargas

Git Bare Repository Worktree Workflow for Claude Code

🚨 CRITICAL: Understanding My Project Structure

I use a bare Git repository approach with Git worktrees extensively. Every subdirectory in my projects represents a different git branch as a worktree.

When you see a project like /home/code/projects/my-app/:

• my-app/ = project container (NOT a working tree)
• my-app/.bare/ = actual Git repository database
• my-app/.git = pointer file directing Git commands to .bare/

ben-vargas

Running AI CLI Tools with Cron for 5-Hour Usage Windows

The Challenge

Anthropic and OpenAI enforce a 5-hour usage window - once I make my first request, I have 5 hours fixed usage before the window expires. This meant:

• ❌ Unpredictable availability - window could expire mid-project or at random times during the work day
• ❌ Frustration when usage windows expire at unopportune times
• ❌ Consistent work schedule and usage window planning

ben-vargas

import { OpenAI } from "openai";
import { runGuardrails } from "@openai/guardrails";
import { z } from "zod";
import { Agent, AgentInputItem, Runner } from "@openai/agents";


// Shared client for guardrails and file search
const client = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });

// Guardrails definitions

ben-vargas

Config for opencode plugin: opencode-openai-codex-auth

For those who might be having edge case behavior, or oddity with start/restart of opencode not retaining last used model properly, or "changing to a different model name" it's likely because of incorrect recommended setup in the opencode.json config file documented in the plugin repo.

You might try the config below, where model id is the key in the models object rather than user-friendly name.

This solved the odd behavior for me with launching opencode and /models slash command - it also provides shorter model names and easier search for all models with "oauth".

ben-vargas

● Complete Testing Results - Clarified for Plugin Normalization

---

GPT-5 Model (via gpt-5-nano)

reasoningEffort	auto	concise	detailed	none
minimal	✅ Success	❌ Unsupported	✅ Success	❌ Invalid
low	✅ Success	❌ Unsupported	✅ Success	❌ Invalid
medium	✅ Success	❌ Unsupported	✅ Success	❌ Invalid

ben-vargas

Configuration Settings - Implementation Plan

This document serves as a comprehensive development guide for implementing configurable settings in the opencode-openai-codex-auth plugin. It outlines the configuration settings that will be supported for all GPT-5 family models when accessed through ChatGPT OAuth (subscription-based) authentication.

Supported Models

This plugin provides access to the following models via ChatGPT Plus/Pro subscription:

1. gpt-5-codex - Optimized for coding tasks with built-in code-aware reasoning
2. gpt-5 - General-purpose reasoning model for complex tasks

ben-vargas

#!/usr/bin/env bash
set -euo pipefail

# ==============================
# Codex Installer Script
# Repository: https://github.com/openai/codex
# ==============================

# Config (overridable via env or flags)
REPO="${REPO:-openai/codex}"

ben-vargas

Tool Streaming Support for AI SDK Provider Codex CLI

Issue Summary

GitHub Issue: #2 - Streaming tool calls? Requested by: @hbmartin Date: January 2025

The user is requesting support for streaming tool calls in streamText and streamObject, similar to the functionality planned for the Claude Code provider. This would enable building UIs that show tool calls and results as they happen.

Important Note on JSON Mode