Using Smart Spectrum GLM4.5 in Codex Configuration

Recently, the frequency of using AI has greatly increased. Copilot runs out of quota every month, and Gemini CLI uses up the Pro quota every day. After running out, I found that I became very unskilled without AI. I started considering getting a "three princesses" setup for myself, so I chose Codex. However, my assistant was reluctant to spend $20 on a membership, so I finally opted for a $20 monthly Coding membership with Zhipu (which is really cheap...). Although Codex supports custom providers, there are still some pitfalls, and there isn't much information online, so I’ll record a simple guide.

Installation#

The installation is very simple and can basically be completed with one command.

npm i -g @openai/codex or
brew install codex or
nix-shell -p codex

config.json Configuration#

The default configuration folder for Codex is in ~/.codex/, which may vary depending on the operating system (Windows might be a bit different). This folder contains the config.json file used for Codex configuration.
First, in version 0.1.2504, many tutorials indicate that configuring config.toml does not take effect, and only json or yaml can be used. There are not many official examples provided for yaml, so it's best to use json.
Here is the official json example:

{
  "model": "o4-mini",
  "provider": "openai",
  "providers": {
    "openai": {
      "name": "OpenAI",
      "baseURL": "https://api.openai.com/v1",
      "envKey": "OPENAI_API_KEY"
    },
    "azure": {
      "name": "AzureOpenAI",
      "baseURL": "https://YOUR_PROJECT_NAME.openai.azure.com/openai",
      "envKey": "AZURE_OPENAI_API_KEY"
    },
    "openrouter": {
      "name": "OpenRouter",
      "baseURL": "https://openrouter.ai/api/v1",
      "envKey": "OPENROUTER_API_KEY"
    },
    "gemini": {
      "name": "Gemini",
      "baseURL": "https://generativelanguage.googleapis.com/v1beta/openai",
      "envKey": "GEMINI_API_KEY"
    },
    "ollama": {
      "name": "Ollama",
      "baseURL": "http://localhost:11434/v1",
      "envKey": "OLLAMA_API_KEY"
    },
    "mistral": {
      "name": "Mistral",
      "baseURL": "https://api.mistral.ai/v1",
      "envKey": "MISTRAL_API_KEY"
    },
    "deepseek": {
      "name": "DeepSeek",
      "baseURL": "https://api.deepseek.com",
      "envKey": "DEEPSEEK_API_KEY"
    },
    "xai": {
      "name": "xAI",
      "baseURL": "https://api.x.ai/v1",
      "envKey": "XAI_API_KEY"
    },
    "groq": {
      "name": "Groq",
      "baseURL": "https://api.groq.com/openai/v1",
      "envKey": "GROQ_API_KEY"
    },
    "arceeai": {
      "name": "ArceeAI",
      "baseURL": "https://conductor.arcee.ai/v1",
      "envKey": "ARCEEAI_API_KEY"
    }
  },
  "history": {
    "maxSize": 1000,
    "saveHistory": true,
    "sensitivePatterns": []
  }
}

In this, the Zhipu provider we need is not included. We can either remove the existing provider or directly add a new provider. The final config.json file is as follows:

{
  "model": "glm-4.5",
  "approvalMode": "suggest",
  "fullAutoErrorMode": "ask-user",
  "notify": true,
  "provider": "zhipuai",
  "providers": {
    "zhipuai": {
      "name": "zhipuai",
      "baseURL": "https://open.bigmodel.cn/api/coding/paas/v4",
      "envKey": "OPENAI_API_KEY"
    }
  },
  "history": {
    "maxSize": 1000,
    "saveHistory": true,
    "sensitivePatterns": []
  }
}

It is important to note that the provider cannot use openai; it must strictly use zhipuai, and the name must also strictly be zhipuai, otherwise, a 404 or 401 error will occur. The envKey means the name of the environment variable to read. It does not actually require you to input the API here. If you configure ZHIPUAI_API_KEY, it will read ZHIPUAI_API_KEY from the environment variables as the API for that provider. This design is intended to switch between different models from different vendors. We only use Zhipu's model, so keeping OPENAI_API_KEY is also fine.

Using in the Project#

Since Codex reads environment variables at startup, as long as you can correctly output the key using echo $OPENAI_API_KEY, it can be used correctly. Codex will also read variables from .env, so there are many configuration methods. The following introduces the mainstream methods.

1. Current Session#

This method only takes effect in the current command line session. When you close the terminal or open a new terminal session, this variable will become invalid.

How to operate

On Linux or macOS, you can use the export command:

export OPENAI_API_KEY='your_key'

On Windows, you can use the set command:

set OPENAI_API_KEY='your_key'

Applicable Scope

Temporary Testing and Debugging: This method is most convenient when you want to quickly test a script or command in the current terminal session without wanting to permanently save the key.
Security Considerations: If your device is shared, or you do not want the key to be stored in a file for a long time, this method can prevent key leakage.

2. `.bashrc` or `.zshrc`#

.bashrc is the configuration file for the Bash shell, and .zshrc is the configuration file for the Zsh shell. These files are automatically executed every time a new terminal session is opened. Adding the key to these files ensures that this environment variable is automatically loaded every time the terminal starts.

How to operate

Open the .bashrc or .zshrc file with a text editor (like nano or vim), and add a line at the end of the file:

export OPENAI_API_KEY='your_key'

After saving and exiting, run source ~/.bashrc or source ~/.zshrc to take effect immediately, or simply close and reopen the terminal.

Applicable Scope

Personal Development Environment: If you are the only user of the device and want to use this key in all terminal sessions, this method is very convenient.
Long-term Use: When you need to frequently use tools that depend on this key, setting it once can make it permanently effective, saving you the trouble of entering it manually each time.

3. `.env` File#

The .env file is a common format for storing environment variables for a project. Many tools and frameworks (including Codex) support automatically loading environment variables from the .env file. This method is usually tied to the project directory.

How to operate

Create a file named .env in your project root directory, and add the following content:

OPENAI_API_KEY='your_key'

Note that there are usually no spaces around the =.

Applicable Scope

Project-Specific Configuration: When your key is only used for a specific project, placing the .env file in the project directory can maintain the independence of the configuration and avoid conflicts with other projects or global environment variables.
Team Collaboration: The .env file can serve as a template, making it easy for team members to configure their own keys. For security, it is usually added to .gitignore to prevent the key from being uploaded to the Git repository.
Framework and Library Support: Many modern development frameworks and libraries (like Next.js, React, Python's dotenv library) natively support reading .env files.

Best Practices:

If you just want to quickly test something, using the current session method is the simplest.
If you frequently use Codex on a personal computer and do not want to set it every time, adding the key to .bashrc is the most suitable.
If you are developing a specific project and want the configuration tied to the project, it is recommended to use the .env file, which is a safer and more universal approach.