Codex Integration Guide

This page explains how to connect Codex through the Encore gateway by using OpenAI-compatible configuration.

Typical Use Cases

Using Codex with config.toml
Using OpenAI-compatible SDKs
Switching Codex traffic to the Encore gateway

Codex Configuration Example

model = "gpt-5.4"
openai_base_url = "https://api.stillhappy.cn/v1"
forced_login_method = "api"

Codex config.toml

For the API key, edit auth.toml:

"OPENAI_API_KEY": YOUR_API_KEY

Codex auth.toml

Important Notes

openai_base_url must include the /v1 suffix
Use forced_login_method = "api" to skip ChatGPT web/OAuth login
Restart Codex completely after editing the config

OpenAI-Compatible Python Example

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.stillhappy.cn/v1",
)

response = client.responses.create(
    model="gpt-5.2-codex",
    input="Write a short Python function that sorts a list.",
)

print(response)

OpenAI-Compatible JavaScript Example

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "YOUR_API_KEY",
  baseURL: "https://api.stillhappy.cn/v1",
});

const response = await client.responses.create({
  model: "gpt-5.2-codex",
  input: "Write a short TypeScript helper to merge arrays.",
});

console.log(response);

Verification

After configuration, verify:

requests are sent to api.stillhappy.cn
the configured model is available through Encore
authentication succeeds with your Encore API key

Latest Model (GPT-5.5) via Codex

If you want to use Codex CLI with GPT-5.5, here's the Windows setup (macOS / Linux are similar).

Windows

1) Install Node.js

Use winget to install the LTS version:

winget install OpenJS.NodeJS.LTS

After installation, verify:

node --version
npm --version

2) Install Codex CLI

In PowerShell or CMD:

npm install -g @openai/codex

Verify:

codex --version

3) Configure Codex

Config directory on Windows:

%USERPROFILE%\.codex

Config file path:

%USERPROFILE%\.codex\config.json

Codex Windows config.json

Write into config.json:

model_provider = "encore___cc"
model = "gpt-5.5"
model_reasoning_effort = "high"
disable_response_storage = true
forced_login_method = "api"

[model_providers.encore___cc]
name = "encore___cc"
base_url = "https://api.stillhappy.cn/v1"
wire_api = "responses"
requires_openai_auth = true

[projects.'c:\users\lenovo']
trust_level = "trusted"

[tui.model_availability_nux]
"gpt-5.5" = 4

[windows]
sandbox = "elevated"

Note: Inside [model_providers.xxx] blocks, the field is base_url (not openai_base_url). The top-level Codex config still uses openai_base_url.

If you can already use other models (like gpt-5.4), you only need to change model = "gpt-5.5" to make GPT-5.5 the default.

Save and restart Codex.

For macOS / Linux, the steps are similar but the config directory is ~/.codex/.

Codex Integration Guide

On this page