Codex Configuration

This repository builds a portable Codex CLI home directory for OpenAI GPT-5.5, DeepSeek, Z.ai coding plan, General Compute, and selected Ollama Cloud coding models.

Requires Codex CLI 0.134.0 or later (profiles use the <name>.config.toml format introduced in 0.134.0).

Configured profiles:

• gpt-5-5
• deepseek-v4-pro
• deepseek-v4-flash
• ollama-cloud-deepseek-v4-pro
• ollama-cloud-deepseek-v4-flash
• zai-coding-glm-5-1
• zai-coding-glm-4-7
• generalcompute-minimax-m2-7
• generalcompute-deepseek-v3-2
• generalcompute-deepseek-v3-1
• ollama-cloud-glm-5-1
• ollama-cloud-glm-5
• ollama-cloud-minimax-m2-7
• ollama-cloud-minimax-m2-5
• ollama-cloud-kimi-k2-6
• ollama-cloud-kimi-k2-5
• ollama-cloud-qwen3-coder-next

Requirements

• Codex CLI >= 0.134.0
• Bash
• jq

Authentication

For OpenAI GPT-5.5, run:

codex login

DeepSeek

DeepSeek models require Moon Bridge running locally. Codex uses the Responses API exclusively; Moon Bridge translates to DeepSeek's Chat Completions API.

1. Install and start Moon Bridge (default: http://127.0.0.1:38440)
2. Configure your DeepSeek API key and DeepSeek routes in Moon Bridge's config.yml
3. Run:

codex --profile deepseek-v4-pro
codex --profile deepseek-v4-flash

Z.AI Coding Plan

Z.AI models require Moon Bridge running locally. Codex uses the Responses API exclusively; Moon Bridge translates to Z.AI's Chat Completions API.

1. Install and start Moon Bridge (default: http://127.0.0.1:38440)
2. Configure Moon Bridge with glm-5.1 / glm-4.7 models, a Z.AI provider using https://api.z.ai/api/coding/paas/v4, and routes for those model names. See docs/moon-bridge-multi-provider.md.
3. Run:

codex --profile zai-coding-glm-5-1
codex --profile zai-coding-glm-4-7

General Compute

General Compute models require Moon Bridge running locally. General Compute provides an OpenAI-compatible Chat Completions API, while Codex uses the Responses API exclusively; Moon Bridge translates between them.

1. Install and start Moon Bridge (default: http://127.0.0.1:38440)
2. Configure Moon Bridge with a General Compute provider using https://api.generalcompute.com/v1, your General Compute API key, and routes for generalcompute-minimax-m2.7, generalcompute-deepseek-v3.2, and generalcompute-deepseek-v3.1. See docs/moon-bridge-generalcompute.md.
3. Run:

codex --profile generalcompute-minimax-m2-7
codex --profile generalcompute-deepseek-v3-2
codex --profile generalcompute-deepseek-v3-1

Ollama Cloud

For Ollama Cloud models, create an API key from ollama.com and set:

export OLLAMA_API_KEY="..."

This configuration talks directly to Ollama Cloud at https://ollama.com/v1; it does not require a local Ollama server proxy.

Wire API Compatibility

Codex CLI uses the Responses API (/v1/responses) exclusively. The older Chat Completions API (/v1/chat/completions) is no longer supported.

Provider	API supported	Works with Codex?	Bridge required?
OpenAI	Responses	Yes	No
Ollama Cloud	Responses (v0.13.3+)	Yes	No
DeepSeek	Chat Completions only	No	Yes — use Moon Bridge
Z.AI	Chat Completions only	No	Yes — use Moon Bridge
General Compute	Chat Completions	No	Yes — use Moon Bridge

Moon Bridge is a Go-based proxy that translates Codex Responses API calls into upstream provider APIs. For DeepSeek plus Z.AI routing, see docs/moon-bridge-multi-provider.md. For General Compute routing, see docs/moon-bridge-generalcompute.md.

Validate

scripts/validate.sh

Install

The installer deletes and recreates ~/.codex. It backs up the existing directory by default.

Preview changes:

./install.sh --dry-run

Install:

./install.sh

Skip the confirmation prompt:

./install.sh --yes

Skip the backup only when you intentionally do not need the current ~/.codex:

./install.sh --yes --no-backup

Usage

codex --profile gpt-5-5
codex --profile deepseek-v4-pro
codex --profile deepseek-v4-flash
codex --profile ollama-cloud-deepseek-v4-pro
codex --profile ollama-cloud-deepseek-v4-flash
codex --profile zai-coding-glm-5-1
codex --profile zai-coding-glm-4-7
codex --profile generalcompute-minimax-m2-7
codex --profile generalcompute-deepseek-v3-2
codex --profile generalcompute-deepseek-v3-1
codex --profile ollama-cloud-glm-5-1
codex --profile ollama-cloud-glm-5
codex --profile ollama-cloud-minimax-m2-7
codex --profile ollama-cloud-minimax-m2-5
codex --profile ollama-cloud-kimi-k2-6
codex --profile ollama-cloud-kimi-k2-5
codex --profile ollama-cloud-qwen3-coder-next

Profile Files

Each profile is a separate <name>.config.toml file in the profiles/ directory. These are copied to ~/.codex/<name>.config.toml during installation. Profile files only contain values that differ from the base config.toml.

Restore Backup

If install created a backup such as ~/.codex.backup.20260421-153000, restore it with:

rm -rf ~/.codex
mv ~/.codex.backup.20260421-153000 ~/.codex

Model Capabilities

Model capability sources and runtime verification status are tracked in docs/model-capabilities.md.

License

This project is licensed under the MIT License. See LICENSE for details.

---

Copyright (c) 2026-present Douglas Hoard