Files
codex/codex-rs/core/src/model_provider_info.rs
Eric Traut 4b9d5c8c1b Add openai_base_url config override for built-in provider (#12031)
We regularly get bug reports from users who mistakenly have the
`OPENAI_BASE_URL` environment variable set. This PR deprecates this
environment variable in favor of a top-level config key
`openai_base_url` that is used for the same purpose. By making it a
config key, it will be more visible to users. It will also participate
in all of the infrastructure we've added for layered and managed
configs.

Summary
- introduce the `openai_base_url` top-level config key, update
schema/tests, and route the built-in openai provider through it while
- fall back to deprecated `OPENAI_BASE_URL` env var but warn user of
deprecation when no `openai_base_url` config key is present
- update CLI, SDK, and TUI code to prefer the new config path (with a
deprecated env-var fallback) and document the SDK behavior change
2026-03-13 20:12:25 -06:00

343 lines
13 KiB
Rust

//! Registry of model providers supported by Codex.
//!
//! Providers can be defined in two places:
//! 1. Built-in defaults compiled into the binary so Codex works out-of-the-box.
//! 2. User-defined entries inside `~/.codex/config.toml` under the `model_providers`
//! key. These override or extend the defaults at runtime.
use crate::auth::AuthMode;
use crate::error::EnvVarError;
use codex_api::Provider as ApiProvider;
use codex_api::provider::RetryConfig as ApiRetryConfig;
use http::HeaderMap;
use http::header::HeaderName;
use http::header::HeaderValue;
use schemars::JsonSchema;
use serde::Deserialize;
use serde::Serialize;
use std::collections::HashMap;
use std::fmt;
use std::time::Duration;
const DEFAULT_STREAM_IDLE_TIMEOUT_MS: u64 = 300_000;
const DEFAULT_STREAM_MAX_RETRIES: u64 = 5;
const DEFAULT_REQUEST_MAX_RETRIES: u64 = 4;
/// Hard cap for user-configured `stream_max_retries`.
const MAX_STREAM_MAX_RETRIES: u64 = 100;
/// Hard cap for user-configured `request_max_retries`.
const MAX_REQUEST_MAX_RETRIES: u64 = 100;
const OPENAI_PROVIDER_NAME: &str = "OpenAI";
pub const OPENAI_PROVIDER_ID: &str = "openai";
const CHAT_WIRE_API_REMOVED_ERROR: &str = "`wire_api = \"chat\"` is no longer supported.\nHow to fix: set `wire_api = \"responses\"` in your provider config.\nMore info: https://github.com/openai/codex/discussions/7782";
pub(crate) const LEGACY_OLLAMA_CHAT_PROVIDER_ID: &str = "ollama-chat";
pub(crate) const OLLAMA_CHAT_PROVIDER_REMOVED_ERROR: &str = "`ollama-chat` is no longer supported.\nHow to fix: replace `ollama-chat` with `ollama` in `model_provider`, `oss_provider`, or `--local-provider`.\nMore info: https://github.com/openai/codex/discussions/7782";
/// Wire protocol that the provider speaks.
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq, Serialize, JsonSchema)]
#[serde(rename_all = "lowercase")]
pub enum WireApi {
/// The Responses API exposed by OpenAI at `/v1/responses`.
#[default]
Responses,
}
impl fmt::Display for WireApi {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
let value = match self {
Self::Responses => "responses",
};
f.write_str(value)
}
}
impl<'de> Deserialize<'de> for WireApi {
fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
where
D: serde::Deserializer<'de>,
{
let value = String::deserialize(deserializer)?;
match value.as_str() {
"responses" => Ok(Self::Responses),
"chat" => Err(serde::de::Error::custom(CHAT_WIRE_API_REMOVED_ERROR)),
_ => Err(serde::de::Error::unknown_variant(&value, &["responses"])),
}
}
}
/// Serializable representation of a provider definition.
#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema)]
#[schemars(deny_unknown_fields)]
pub struct ModelProviderInfo {
/// Friendly display name.
pub name: String,
/// Base URL for the provider's OpenAI-compatible API.
pub base_url: Option<String>,
/// Environment variable that stores the user's API key for this provider.
pub env_key: Option<String>,
/// Optional instructions to help the user get a valid value for the
/// variable and set it.
pub env_key_instructions: Option<String>,
/// Value to use with `Authorization: Bearer <token>` header. Use of this
/// config is discouraged in favor of `env_key` for security reasons, but
/// this may be necessary when using this programmatically.
pub experimental_bearer_token: Option<String>,
/// Which wire protocol this provider expects.
#[serde(default)]
pub wire_api: WireApi,
/// Optional query parameters to append to the base URL.
pub query_params: Option<HashMap<String, String>>,
/// Additional HTTP headers to include in requests to this provider where
/// the (key, value) pairs are the header name and value.
pub http_headers: Option<HashMap<String, String>>,
/// Optional HTTP headers to include in requests to this provider where the
/// (key, value) pairs are the header name and _environment variable_ whose
/// value should be used. If the environment variable is not set, or the
/// value is empty, the header will not be included in the request.
pub env_http_headers: Option<HashMap<String, String>>,
/// Maximum number of times to retry a failed HTTP request to this provider.
pub request_max_retries: Option<u64>,
/// Number of times to retry reconnecting a dropped streaming response before failing.
pub stream_max_retries: Option<u64>,
/// Idle timeout (in milliseconds) to wait for activity on a streaming response before treating
/// the connection as lost.
pub stream_idle_timeout_ms: Option<u64>,
/// Does this provider require an OpenAI API Key or ChatGPT login token? If true,
/// user is presented with login screen on first run, and login preference and token/key
/// are stored in auth.json. If false (which is the default), login screen is skipped,
/// and API key (if needed) comes from the "env_key" environment variable.
#[serde(default)]
pub requires_openai_auth: bool,
/// Whether this provider supports the Responses API WebSocket transport.
#[serde(default)]
pub supports_websockets: bool,
}
impl ModelProviderInfo {
fn build_header_map(&self) -> crate::error::Result<HeaderMap> {
let capacity = self.http_headers.as_ref().map_or(0, HashMap::len)
+ self.env_http_headers.as_ref().map_or(0, HashMap::len);
let mut headers = HeaderMap::with_capacity(capacity);
if let Some(extra) = &self.http_headers {
for (k, v) in extra {
if let (Ok(name), Ok(value)) = (HeaderName::try_from(k), HeaderValue::try_from(v)) {
headers.insert(name, value);
}
}
}
if let Some(env_headers) = &self.env_http_headers {
for (header, env_var) in env_headers {
if let Ok(val) = std::env::var(env_var)
&& !val.trim().is_empty()
&& let (Ok(name), Ok(value)) =
(HeaderName::try_from(header), HeaderValue::try_from(val))
{
headers.insert(name, value);
}
}
}
Ok(headers)
}
pub(crate) fn to_api_provider(
&self,
auth_mode: Option<AuthMode>,
) -> crate::error::Result<ApiProvider> {
let default_base_url = if matches!(auth_mode, Some(AuthMode::Chatgpt)) {
"https://chatgpt.com/backend-api/codex"
} else {
"https://api.openai.com/v1"
};
let base_url = self
.base_url
.clone()
.unwrap_or_else(|| default_base_url.to_string());
let headers = self.build_header_map()?;
let retry = ApiRetryConfig {
max_attempts: self.request_max_retries(),
base_delay: Duration::from_millis(200),
retry_429: false,
retry_5xx: true,
retry_transport: true,
};
Ok(ApiProvider {
name: self.name.clone(),
base_url,
query_params: self.query_params.clone(),
headers,
retry,
stream_idle_timeout: self.stream_idle_timeout(),
})
}
/// If `env_key` is Some, returns the API key for this provider if present
/// (and non-empty) in the environment. If `env_key` is required but
/// cannot be found, returns an error.
pub fn api_key(&self) -> crate::error::Result<Option<String>> {
match &self.env_key {
Some(env_key) => {
let api_key = std::env::var(env_key)
.ok()
.filter(|v| !v.trim().is_empty())
.ok_or_else(|| {
crate::error::CodexErr::EnvVar(EnvVarError {
var: env_key.clone(),
instructions: self.env_key_instructions.clone(),
})
})?;
Ok(Some(api_key))
}
None => Ok(None),
}
}
/// Effective maximum number of request retries for this provider.
pub fn request_max_retries(&self) -> u64 {
self.request_max_retries
.unwrap_or(DEFAULT_REQUEST_MAX_RETRIES)
.min(MAX_REQUEST_MAX_RETRIES)
}
/// Effective maximum number of stream reconnection attempts for this provider.
pub fn stream_max_retries(&self) -> u64 {
self.stream_max_retries
.unwrap_or(DEFAULT_STREAM_MAX_RETRIES)
.min(MAX_STREAM_MAX_RETRIES)
}
/// Effective idle timeout for streaming responses.
pub fn stream_idle_timeout(&self) -> Duration {
self.stream_idle_timeout_ms
.map(Duration::from_millis)
.unwrap_or(Duration::from_millis(DEFAULT_STREAM_IDLE_TIMEOUT_MS))
}
pub fn create_openai_provider(base_url: Option<String>) -> ModelProviderInfo {
ModelProviderInfo {
name: OPENAI_PROVIDER_NAME.into(),
base_url,
env_key: None,
env_key_instructions: None,
experimental_bearer_token: None,
wire_api: WireApi::Responses,
query_params: None,
http_headers: Some(
[("version".to_string(), env!("CARGO_PKG_VERSION").to_string())]
.into_iter()
.collect(),
),
env_http_headers: Some(
[
(
"OpenAI-Organization".to_string(),
"OPENAI_ORGANIZATION".to_string(),
),
("OpenAI-Project".to_string(), "OPENAI_PROJECT".to_string()),
]
.into_iter()
.collect(),
),
// Use global defaults for retry/timeout unless overridden in config.toml.
request_max_retries: None,
stream_max_retries: None,
stream_idle_timeout_ms: None,
requires_openai_auth: true,
supports_websockets: true,
}
}
pub fn is_openai(&self) -> bool {
self.name == OPENAI_PROVIDER_NAME
}
}
pub const DEFAULT_LMSTUDIO_PORT: u16 = 1234;
pub const DEFAULT_OLLAMA_PORT: u16 = 11434;
pub const LMSTUDIO_OSS_PROVIDER_ID: &str = "lmstudio";
pub const OLLAMA_OSS_PROVIDER_ID: &str = "ollama";
/// Built-in default provider list.
pub fn built_in_model_providers(
openai_base_url: Option<String>,
) -> HashMap<String, ModelProviderInfo> {
use ModelProviderInfo as P;
let openai_provider = P::create_openai_provider(openai_base_url);
// We do not want to be in the business of adjucating which third-party
// providers are bundled with Codex CLI, so we only include the OpenAI and
// open source ("oss") providers by default. Users are encouraged to add to
// `model_providers` in config.toml to add their own providers.
[
(OPENAI_PROVIDER_ID, openai_provider),
(
OLLAMA_OSS_PROVIDER_ID,
create_oss_provider(DEFAULT_OLLAMA_PORT, WireApi::Responses),
),
(
LMSTUDIO_OSS_PROVIDER_ID,
create_oss_provider(DEFAULT_LMSTUDIO_PORT, WireApi::Responses),
),
]
.into_iter()
.map(|(k, v)| (k.to_string(), v))
.collect()
}
pub fn create_oss_provider(default_provider_port: u16, wire_api: WireApi) -> ModelProviderInfo {
// These CODEX_OSS_ environment variables are experimental: we may
// switch to reading values from config.toml instead.
let default_codex_oss_base_url = format!(
"http://localhost:{codex_oss_port}/v1",
codex_oss_port = std::env::var("CODEX_OSS_PORT")
.ok()
.filter(|value| !value.trim().is_empty())
.and_then(|value| value.parse::<u16>().ok())
.unwrap_or(default_provider_port)
);
let codex_oss_base_url = std::env::var("CODEX_OSS_BASE_URL")
.ok()
.filter(|v| !v.trim().is_empty())
.unwrap_or(default_codex_oss_base_url);
create_oss_provider_with_base_url(&codex_oss_base_url, wire_api)
}
pub fn create_oss_provider_with_base_url(base_url: &str, wire_api: WireApi) -> ModelProviderInfo {
ModelProviderInfo {
name: "gpt-oss".into(),
base_url: Some(base_url.into()),
env_key: None,
env_key_instructions: None,
experimental_bearer_token: None,
wire_api,
query_params: None,
http_headers: None,
env_http_headers: None,
request_max_retries: None,
stream_max_retries: None,
stream_idle_timeout_ms: None,
requires_openai_auth: false,
supports_websockets: false,
}
}
#[cfg(test)]
#[path = "model_provider_info_tests.rs"]
mod tests;