agent/codex
2
0
Fork 0
mirror of https://github.com/pchuan98/codex.git synced 2026-07-01 00:31:56 +08:00
Code Issues Packages Projects Releases Wiki Activity

Remove legacy ModelInfo and merge it with ModelFamily (#7748)

Browse Source 
This is a step towards removing the need to know `model` when
constructing config. We firstly don't need to know `model_info` and just
respect if the user has already set it. Next step, we don't need to know
`model` unless the user explicitly set it in `config.toml`
This commit is contained in:
Ahmed Ibrahim
2025-12-08 15:29:37 -08:00
committed by GitHub
parent ac5fa6baf8
commit 382f047a10
10 changed files with 205 additions and 225 deletions
Download Patch File Download Diff File Expand all files Collapse all files
codex-rs/core/src/client.rs
+2 -11
View File
@@ -48,7 +48,6 @@ use crate::error::Result;
use crate::flags::CODEX_RS_SSE_FIXTURE;
use crate::model_provider_info::ModelProviderInfo;
use crate::model_provider_info::WireApi;
use crate::openai_model_info::get_model_info;
use crate::openai_models::model_family::ModelFamily;
use crate::tools::spec::create_tools_json_for_chat_completions_api;
use crate::tools::spec::create_tools_json_for_responses_api;
@@ -95,19 +94,11 @@ impl ModelClient {
pub fn get_model_context_window(&self) -> Option<i64> {
let model_family = self.get_model_family();
let effective_context_window_percent = model_family.effective_context_window_percent;
self.config
.model_context_window
.or_else(|| get_model_info(&model_family).map(|info| info.context_window))
model_family
.context_window
.map(|w| w.saturating_mul(effective_context_window_percent) / 100)
}
pub fn get_auto_compact_token_limit(&self) -> Option<i64> {
let model_family = self.get_model_family();
self.config.model_auto_compact_token_limit.or_else(|| {
get_model_info(&model_family).and_then(|info| info.auto_compact_token_limit)
})
}
pub fn config(&self) -> Arc<Config> {
Arc::clone(&self.config)
}
codex-rs/core/src/codex.rs
+3 -10
View File
@@ -80,7 +80,6 @@ use crate::exec::StreamOutput;
use crate::exec_policy::ExecPolicyUpdateError;
use crate::mcp::auth::compute_auth_statuses;
use crate::mcp_connection_manager::McpConnectionManager;
use crate::openai_model_info::get_model_info;
use crate::project_doc::get_user_instructions;
use crate::protocol::AgentMessageContentDeltaEvent;
use crate::protocol::AgentReasoningSectionBreakEvent;
@@ -415,15 +414,11 @@ impl Session {
otel_event_manager: &OtelEventManager,
provider: ModelProviderInfo,
session_configuration: &SessionConfiguration,
mut per_turn_config: Config,
per_turn_config: Config,
model_family: ModelFamily,
conversation_id: ConversationId,
sub_id: String,
) -> TurnContext {
if let Some(model_info) = get_model_info(&model_family) {
per_turn_config.model_context_window = Some(model_info.context_window);
}
let otel_event_manager = otel_event_manager.clone().with_model(
session_configuration.model.as_str(),
model_family.slug.as_str(),
@@ -1955,9 +1950,6 @@ async fn spawn_review_thread(
per_turn_config.model_reasoning_effort = Some(ReasoningEffortConfig::Low);
per_turn_config.model_reasoning_summary = ReasoningSummaryConfig::Detailed;
per_turn_config.features = review_features.clone();
if let Some(model_info) = get_model_info(&model_family) {
per_turn_config.model_context_window = Some(model_info.context_window);
}
let otel_event_manager = parent_turn_context
.client
@@ -2097,7 +2089,8 @@ pub(crate) async fn run_task(
} = turn_output;
let limit = turn_context
.client
.get_auto_compact_token_limit()
.get_model_family()
.auto_compact_token_limit()
.unwrap_or(i64::MAX);
let total_usage_tokens = sess.get_total_token_usage().await;
let token_limit_reached = total_usage_tokens >= limit;
codex-rs/core/src/config/mod.rs
+11 -24
View File
@@ -26,8 +26,6 @@ use crate::model_provider_info::LMSTUDIO_OSS_PROVIDER_ID;
use crate::model_provider_info::ModelProviderInfo;
use crate::model_provider_info::OLLAMA_OSS_PROVIDER_ID;
use crate::model_provider_info::built_in_model_providers;
use crate::openai_model_info::get_model_info;
use crate::openai_models::model_family::find_family_for_model;
use crate::project_doc::DEFAULT_PROJECT_DOC_FILENAME;
use crate::project_doc::LOCAL_PROJECT_DOC_FILENAME;
use crate::protocol::AskForApproval;
@@ -1106,23 +1104,12 @@ impl Config {
let forced_login_method = cfg.forced_login_method;
// todo(aibrahim): make model optional
let model = model
.or(config_profile.model)
.or(cfg.model)
.unwrap_or_else(default_model);
let model_family = find_family_for_model(&model);
let openai_model_info = get_model_info(&model_family);
let model_context_window = cfg
.model_context_window
.or_else(|| openai_model_info.as_ref().map(|info| info.context_window));
let model_auto_compact_token_limit = cfg.model_auto_compact_token_limit.or_else(|| {
openai_model_info
.as_ref()
.and_then(|info| info.auto_compact_token_limit)
});
let compact_prompt = compact_prompt.or(cfg.compact_prompt).and_then(|value| {
let trimmed = value.trim();
if trimmed.is_empty() {
@@ -1168,8 +1155,8 @@ impl Config {
let config = Self {
model,
review_model,
model_context_window,
model_auto_compact_token_limit,
model_context_window: cfg.model_context_window,
model_auto_compact_token_limit: cfg.model_auto_compact_token_limit,
model_provider_id,
model_provider,
cwd: resolved_cwd,
@@ -2950,8 +2937,8 @@ model_verbosity = "high"
Config {
model: "o3".to_string(),
review_model: OPENAI_DEFAULT_REVIEW_MODEL.to_string(),
model_context_window: Some(200_000),
model_auto_compact_token_limit: Some(180_000),
model_context_window: None,
model_auto_compact_token_limit: None,
model_provider_id: "openai".to_string(),
model_provider: fixture.openai_provider.clone(),
approval_policy: AskForApproval::Never,
@@ -3025,8 +3012,8 @@ model_verbosity = "high"
let expected_gpt3_profile_config = Config {
model: "gpt-3.5-turbo".to_string(),
review_model: OPENAI_DEFAULT_REVIEW_MODEL.to_string(),
model_context_window: Some(16_385),
model_auto_compact_token_limit: Some(14_746),
model_context_window: None,
model_auto_compact_token_limit: None,
model_provider_id: "openai-chat-completions".to_string(),
model_provider: fixture.openai_chat_completions_provider.clone(),
approval_policy: AskForApproval::UnlessTrusted,
@@ -3115,8 +3102,8 @@ model_verbosity = "high"
let expected_zdr_profile_config = Config {
model: "o3".to_string(),
review_model: OPENAI_DEFAULT_REVIEW_MODEL.to_string(),
model_context_window: Some(200_000),
model_auto_compact_token_limit: Some(180_000),
model_context_window: None,
model_auto_compact_token_limit: None,
model_provider_id: "openai".to_string(),
model_provider: fixture.openai_provider.clone(),
approval_policy: AskForApproval::OnFailure,
@@ -3191,8 +3178,8 @@ model_verbosity = "high"
let expected_gpt5_profile_config = Config {
model: "gpt-5.1".to_string(),
review_model: OPENAI_DEFAULT_REVIEW_MODEL.to_string(),
model_context_window: Some(272_000),
model_auto_compact_token_limit: Some(244_800),
model_context_window: None,
model_auto_compact_token_limit: None,
model_provider_id: "openai".to_string(),
model_provider: fixture.openai_provider.clone(),
approval_policy: AskForApproval::OnFailure,
codex-rs/core/src/lib.rs
-1
View File
@@ -67,7 +67,6 @@ pub use conversation_manager::NewConversation;
pub use auth::AuthManager;
pub use auth::CodexAuth;
pub mod default_client;
mod openai_model_info;
pub mod project_doc;
mod rollout;
pub(crate) mod safety;
codex-rs/core/src/openai_model_info.rs
-83
View File
@@ -1,83 +0,0 @@
use crate::openai_models::model_family::ModelFamily;
// Shared constants for commonly used window/token sizes.
pub(crate) const CONTEXT_WINDOW_272K: i64 = 272_000;
/// Metadata about a model, particularly OpenAI models.
/// We may want to consider including details like the pricing for
/// input tokens, output tokens, etc., though users will need to be able to
/// override this in config.toml, as this information can get out of date.
/// Though this would help present more accurate pricing information in the UI.
#[derive(Debug)]
pub(crate) struct ModelInfo {
/// Size of the context window in tokens. This is the maximum size of the input context.
pub(crate) context_window: i64,
/// Token threshold where we should automatically compact conversation history. This considers
/// input tokens + output tokens of this turn.
pub(crate) auto_compact_token_limit: Option<i64>,
}
impl ModelInfo {
const fn new(context_window: i64) -> Self {
Self {
context_window,
auto_compact_token_limit: Some(Self::default_auto_compact_limit(context_window)),
}
}
const fn default_auto_compact_limit(context_window: i64) -> i64 {
(context_window * 9) / 10
}
}
pub(crate) fn get_model_info(model_family: &ModelFamily) -> Option<ModelInfo> {
let slug = model_family.slug.as_str();
match slug {
// OSS models have a 128k shared token pool.
// Arbitrarily splitting it: 3/4 input context, 1/4 output.
// https://openai.com/index/gpt-oss-model-card/
"gpt-oss-20b" => Some(ModelInfo::new(96_000)),
"gpt-oss-120b" => Some(ModelInfo::new(96_000)),
// https://platform.openai.com/docs/models/o3
"o3" => Some(ModelInfo::new(200_000)),
// https://platform.openai.com/docs/models/o4-mini
"o4-mini" => Some(ModelInfo::new(200_000)),
// https://platform.openai.com/docs/models/codex-mini-latest
"codex-mini-latest" => Some(ModelInfo::new(200_000)),
// As of Jun 25, 2025, gpt-4.1 defaults to gpt-4.1-2025-04-14.
// https://platform.openai.com/docs/models/gpt-4.1
"gpt-4.1" | "gpt-4.1-2025-04-14" => Some(ModelInfo::new(1_047_576)),
// As of Jun 25, 2025, gpt-4o defaults to gpt-4o-2024-08-06.
// https://platform.openai.com/docs/models/gpt-4o
"gpt-4o" | "gpt-4o-2024-08-06" => Some(ModelInfo::new(128_000)),
// https://platform.openai.com/docs/models/gpt-4o?snapshot=gpt-4o-2024-05-13
"gpt-4o-2024-05-13" => Some(ModelInfo::new(128_000)),
// https://platform.openai.com/docs/models/gpt-4o?snapshot=gpt-4o-2024-11-20
"gpt-4o-2024-11-20" => Some(ModelInfo::new(128_000)),
// https://platform.openai.com/docs/models/gpt-3.5-turbo
"gpt-3.5-turbo" => Some(ModelInfo::new(16_385)),
_ if slug.starts_with("gpt-5-codex")
|| slug.starts_with("gpt-5.1-codex")
|| slug.starts_with("gpt-5.1-codex-max") =>
{
Some(ModelInfo::new(CONTEXT_WINDOW_272K))
}
_ if slug.starts_with("gpt-5") => Some(ModelInfo::new(CONTEXT_WINDOW_272K)),
_ if slug.starts_with("codex-") => Some(ModelInfo::new(CONTEXT_WINDOW_272K)),
_ if slug.starts_with("exp-") => Some(ModelInfo::new(CONTEXT_WINDOW_272K)),
_ => None,
}
}
codex-rs/core/src/openai_models/model_family.rs
+53 -5
View File
@@ -15,6 +15,7 @@ const BASE_INSTRUCTIONS: &str = include_str!("../../prompt.md");
const GPT_5_CODEX_INSTRUCTIONS: &str = include_str!("../../gpt_5_codex_prompt.md");
const GPT_5_1_INSTRUCTIONS: &str = include_str!("../../gpt_5_1_prompt.md");
const GPT_5_1_CODEX_MAX_INSTRUCTIONS: &str = include_str!("../../gpt-5.1-codex-max_prompt.md");
pub(crate) const CONTEXT_WINDOW_272K: i64 = 272_000;
/// A model family is a group of models that share certain characteristics.
#[derive(Debug, Clone, PartialEq, Eq, Hash)]
@@ -23,14 +24,20 @@ pub struct ModelFamily {
/// "gpt-4.1-2025-04-14".
pub slug: String,
/// The model family name, e.g. "gpt-4.1". Note this should able to be used
/// with [`crate::openai_model_info::get_model_info`].
/// The model family name, e.g. "gpt-4.1". This string is used when deriving
/// default metadata for the family, such as context windows.
pub family: String,
/// True if the model needs additional instructions on how to use the
/// "virtual" `apply_patch` CLI.
pub needs_special_apply_patch_instructions: bool,
/// Maximum supported context window, if known.
pub context_window: Option<i64>,
/// Token threshold for automatic compaction if config does not override it.
auto_compact_token_limit: Option<i64>,
// Whether the `reasoning` field can be set when making a request to this
// model family. Note it has `effort` and `summary` subfields (though
// `summary` is optional).
@@ -82,6 +89,12 @@ impl ModelFamily {
if let Some(reasoning_summary_format) = config.model_reasoning_summary_format.as_ref() {
self.reasoning_summary_format = reasoning_summary_format.clone();
}
if let Some(context_window) = config.model_context_window {
self.context_window = Some(context_window);
}
if let Some(auto_compact_token_limit) = config.model_auto_compact_token_limit {
self.auto_compact_token_limit = Some(auto_compact_token_limit);
}
self
}
pub fn with_remote_overrides(mut self, remote_models: Vec<ModelInfo>) -> Self {
@@ -93,6 +106,15 @@ impl ModelFamily {
}
self
}
pub fn auto_compact_token_limit(&self) -> Option<i64> {
self.auto_compact_token_limit
.or(self.context_window.map(Self::default_auto_compact_limit))
}
const fn default_auto_compact_limit(context_window: i64) -> i64 {
(context_window * 9) / 10
}
}
macro_rules! model_family {
@@ -105,6 +127,8 @@ macro_rules! model_family {
slug: $slug.to_string(),
family: $family.to_string(),
needs_special_apply_patch_instructions: false,
context_window: Some(CONTEXT_WINDOW_272K),
auto_compact_token_limit: None,
supports_reasoning_summaries: false,
reasoning_summary_format: ReasoningSummaryFormat::None,
supports_parallel_tool_calls: false,
@@ -136,12 +160,14 @@ pub fn find_family_for_model(slug: &str) -> ModelFamily {
slug, "o3",
supports_reasoning_summaries: true,
needs_special_apply_patch_instructions: true,
context_window: Some(200_000),
)
} else if slug.starts_with("o4-mini") {
model_family!(
slug, "o4-mini",
supports_reasoning_summaries: true,
needs_special_apply_patch_instructions: true,
context_window: Some(200_000),
)
} else if slug.starts_with("codex-mini-latest") {
model_family!(
@@ -149,18 +175,32 @@ pub fn find_family_for_model(slug: &str) -> ModelFamily {
supports_reasoning_summaries: true,
needs_special_apply_patch_instructions: true,
shell_type: ConfigShellToolType::Local,
context_window: Some(200_000),
)
} else if slug.starts_with("gpt-4.1") {
model_family!(
slug, "gpt-4.1",
needs_special_apply_patch_instructions: true,
context_window: Some(1_047_576),
)
} else if slug.starts_with("gpt-oss") || slug.starts_with("openai/gpt-oss") {
model_family!(slug, "gpt-oss", apply_patch_tool_type: Some(ApplyPatchToolType::Function))
model_family!(
slug, "gpt-oss",
apply_patch_tool_type: Some(ApplyPatchToolType::Function),
context_window: Some(96_000),
)
} else if slug.starts_with("gpt-4o") {
model_family!(slug, "gpt-4o", needs_special_apply_patch_instructions: true)
model_family!(
slug, "gpt-4o",
needs_special_apply_patch_instructions: true,
context_window: Some(128_000),
)
} else if slug.starts_with("gpt-3.5") {
model_family!(slug, "gpt-3.5", needs_special_apply_patch_instructions: true)
model_family!(
slug, "gpt-3.5",
needs_special_apply_patch_instructions: true,
context_window: Some(16_385),
)
} else if slug.starts_with("test-gpt-5") {
model_family!(
slug, slug,
@@ -196,6 +236,7 @@ pub fn find_family_for_model(slug: &str) -> ModelFamily {
supports_parallel_tool_calls: true,
support_verbosity: true,
truncation_policy: TruncationPolicy::Tokens(10_000),
context_window: Some(CONTEXT_WINDOW_272K),
)
} else if slug.starts_with("exp-") {
model_family!(
@@ -209,6 +250,7 @@ pub fn find_family_for_model(slug: &str) -> ModelFamily {
truncation_policy: TruncationPolicy::Bytes(10_000),
shell_type: ConfigShellToolType::UnifiedExec,
supports_parallel_tool_calls: true,
context_window: Some(CONTEXT_WINDOW_272K),
)
// Production models.
@@ -223,6 +265,7 @@ pub fn find_family_for_model(slug: &str) -> ModelFamily {
supports_parallel_tool_calls: true,
support_verbosity: false,
truncation_policy: TruncationPolicy::Tokens(10_000),
context_window: Some(CONTEXT_WINDOW_272K),
)
} else if slug.starts_with("gpt-5-codex")
|| slug.starts_with("gpt-5.1-codex")
@@ -238,6 +281,7 @@ pub fn find_family_for_model(slug: &str) -> ModelFamily {
supports_parallel_tool_calls: true,
support_verbosity: false,
truncation_policy: TruncationPolicy::Tokens(10_000),
context_window: Some(CONTEXT_WINDOW_272K),
)
} else if slug.starts_with("gpt-5.1") {
model_family!(
@@ -251,6 +295,7 @@ pub fn find_family_for_model(slug: &str) -> ModelFamily {
truncation_policy: TruncationPolicy::Bytes(10_000),
shell_type: ConfigShellToolType::ShellCommand,
supports_parallel_tool_calls: true,
context_window: Some(CONTEXT_WINDOW_272K),
)
} else if slug.starts_with("gpt-5") {
model_family!(
@@ -260,6 +305,7 @@ pub fn find_family_for_model(slug: &str) -> ModelFamily {
shell_type: ConfigShellToolType::Default,
support_verbosity: true,
truncation_policy: TruncationPolicy::Bytes(10_000),
context_window: Some(CONTEXT_WINDOW_272K),
)
} else {
derive_default_model_family(slug)
@@ -271,6 +317,8 @@ fn derive_default_model_family(model: &str) -> ModelFamily {
slug: model.to_string(),
family: model.to_string(),
needs_special_apply_patch_instructions: false,
context_window: None,
auto_compact_token_limit: None,
supports_reasoning_summaries: false,
reasoning_summary_format: ReasoningSummaryFormat::None,
supports_parallel_tool_calls: false,