下载 Codex App，登录后将配置写入 config.toml，几步完成 Flame API 接入。
下载并打开 Codex App，输入 API Key 完成登录。
进入设置里的配置页面，点击 Open config.toml 打开本地配置文件。
根据系统选择对应配置块，复制到 config.toml 后保存。
model_provider = "Flame"
model = "gpt-5.5"
model_reasoning_effort = "medium"
network_access = "enabled"
disable_response_storage = true
windows_wsl_setup_acknowledged = true
model_verbosity = "medium"
personality = "friendly"
[model_providers.Flame]
name = "Flame"
base_url = "https://api.flameapi.cn/v1"
wire_api = "responses"
requires_openai_auth = true
[windows]
sandbox = "elevated"model_provider = "Flame"
model = "gpt-5.5"
model_reasoning_effort = "medium"
network_access = "enabled"
disable_response_storage = true
windows_wsl_setup_acknowledged = true
model_verbosity = "medium"
personality = "friendly"
[model_providers.Flame]
name = "Flame"
base_url = "https://api.flameapi.cn/v1"
wire_api = "responses"
requires_openai_auth = true下表对应配置项、取值和用途说明。
| 参数 | 值 | 说明 |
|---|---|---|
| model_provider | "Flame" | 使用 Flame API 提供商 |
| model | "gpt-5.5" | 使用的模型名称 |
| model_reasoning_effort | "medium / high" | 推理强度,设为平衡型 / 高质量模式 |
| network_access | "enabled" | 启用网络访问 |
| disable_response_storage | TRUE | 不在服务端存储响应内容 |
| windows_wsl_setup_acknowledged | TRUE | 确认 WSL 设置（Windows 用户需开启） |
| model_verbosity | "high" | 输出详细程度,设为详细 |
| personality | "friendly" | AI 风格,设为友好模式 |
| base_url | "https://api.flameapi.cn/v1" | API 接入地址 |
| wire_api | "responses" | 使用 Responses API 协议 |
| requires_openai_auth | TRUE | 使用 OpenAI 兼容鉴权方式 |