OpenAICodexConfig

OpenAICodexConfig is the config used by OpenAICodexClient::new(...) and stored in AppConfig as openai_codex.

Definition

pub struct OpenAICodexConfig {
    pub access_token: Option<String>,
    pub base_url: String,
    pub model: String,
    pub pricing_model: Option<String>,
    pub max_output_tokens: Option<i32>,
    pub temperature: Option<f32>,
    pub top_p: Option<f32>,
    pub stream: bool,
    pub reasoning: Option<ReasoningConfig>,
    pub text_verbosity: Option<TextVerbosity>,
    pub retry: Option<RetryConfig>,
    pub auth_file: PathBuf,
    pub originator: String,
}

Defaults from the current code

• base_url = "https://chatgpt.com/backend-api"
• model = "gpt-5.4"
• max_output_tokens = Some(4096)
• stream = true
• retry = Some(RetryConfig::default())
• auth_file = ~/.appam/auth.json
• originator = "pi"

Authentication precedence

Runtime authentication is resolved in this order:

1. OpenAICodexConfig.access_token
2. OPENAI_CODEX_ACCESS_TOKEN
3. cached OAuth credentials in auth_file

The auth-cache path defaults to ~/.appam/auth.json and can be overridden with OPENAI_CODEX_AUTH_FILE.

Validation behavior

validate() currently enforces:

• temperature in 0.0..=2.0
• top_p in 0.0..=1.0
• reasoning summaries are rejected when the resolved effort becomes none
• sampling compatibility is checked against the normalized Codex model and resolved reasoning effort

Builder hooks

AgentBuilder exposes the current Codex-relevant overrides:

.openai_codex_access_token(...)
.openai_reasoning(...)
.openai_text_verbosity(...)

Shared runtime overrides such as .model(...), .max_tokens(...), .temperature(...), .top_p(...), and .retry(...) also flow into the Codex config.