Override truncation policy at model info level (#8856)

We used to override truncation policy by comparing model info vs config
value in context manager. A better way to do it is to construct model
info using the config value

codex-rs/core/src/codex.rs

@@ -544,10 +544,7 @@ impl Session {
             final_output_json_schema: None,
             codex_linux_sandbox_exe: per_turn_config.codex_linux_sandbox_exe.clone(),
             tool_call_gate: Arc::new(ReadinessFlag::new()),
-            truncation_policy: TruncationPolicy::new(
-                per_turn_config.as_ref(),
-                model_info.truncation_policy.into(),
-            ),
+            truncation_policy: model_info.truncation_policy.into(),
         }
     }
 
@@ -2244,10 +2241,7 @@ async fn spawn_review_thread(
         final_output_json_schema: None,
         codex_linux_sandbox_exe: parent_turn_context.codex_linux_sandbox_exe.clone(),
         tool_call_gate: Arc::new(ReadinessFlag::new()),
-        truncation_policy: TruncationPolicy::new(
-            &per_turn_config,
-            model_info.truncation_policy.into(),
-        ),
+        truncation_policy: model_info.truncation_policy.into(),
     };
 
     // Seed the child task with the review prompt as the initial user message.

codex-rs/core/src/config/mod.rs

@@ -268,7 +268,6 @@ pub struct Config {
     /// Additional filenames to try when looking for project-level docs.
     pub project_doc_fallback_filenames: Vec<String>,
 
-    // todo(aibrahim): this should be used in the override model info
     /// Token budget applied when storing tool/function outputs in the context manager.
     pub tool_output_token_limit: Option<usize>,
 

codex-rs/core/src/models_manager/model_info.rs

@@ -5,9 +5,11 @@ use codex_protocol::openai_models::ModelInfo;
 use codex_protocol::openai_models::ModelVisibility;
 use codex_protocol::openai_models::ReasoningEffort;
 use codex_protocol::openai_models::ReasoningEffortPreset;
+use codex_protocol::openai_models::TruncationMode;
 use codex_protocol::openai_models::TruncationPolicyConfig;
 
 use crate::config::Config;
+use crate::truncate::approx_bytes_for_tokens;
 use tracing::warn;
 
 const BASE_INSTRUCTIONS: &str = include_str!("../../prompt.md");
@@ -71,6 +73,19 @@ pub(crate) fn with_config_overrides(mut model: ModelInfo, config: &Config) -> Mo
     if let Some(auto_compact_token_limit) = config.model_auto_compact_token_limit {
         model.auto_compact_token_limit = Some(auto_compact_token_limit);
     }
+    if let Some(token_limit) = config.tool_output_token_limit {
+        model.truncation_policy = match model.truncation_policy.mode {
+            TruncationMode::Bytes => {
+                let byte_limit =
+                    i64::try_from(approx_bytes_for_tokens(token_limit)).unwrap_or(i64::MAX);
+                TruncationPolicyConfig::bytes(byte_limit)
+            }
+            TruncationMode::Tokens => {
+                let limit = i64::try_from(token_limit).unwrap_or(i64::MAX);
+                TruncationPolicyConfig::tokens(limit)
+            }
+        };
+    }
     model
 }
 

codex-rs/core/src/truncate.rs

@@ -2,7 +2,6 @@
 //! and suffix on UTF-8 boundaries, and helpers for line/token‑based truncation
 //! used across the core crate.
 
-use crate::config::Config;
 use codex_protocol::models::FunctionCallOutputContentItem;
 use codex_protocol::openai_models::TruncationMode;
 use codex_protocol::openai_models::TruncationPolicyConfig;
@@ -47,27 +46,6 @@ impl TruncationPolicy {
         }
     }
 
-    pub fn new(config: &Config, truncation_policy: TruncationPolicy) -> Self {
-        let config_token_limit = config.tool_output_token_limit;
-
-        match truncation_policy {
-            TruncationPolicy::Bytes(family_bytes) => {
-                if let Some(token_limit) = config_token_limit {
-                    Self::Bytes(approx_bytes_for_tokens(token_limit))
-                } else {
-                    Self::Bytes(family_bytes)
-                }
-            }
-            TruncationPolicy::Tokens(family_tokens) => {
-                if let Some(token_limit) = config_token_limit {
-                    Self::Tokens(token_limit)
-                } else {
-                    Self::Tokens(family_tokens)
-                }
-            }
-        }
-    }
-
     /// Returns a token budget derived from this policy.
     ///
     /// - For `Tokens`, this is the explicit token limit.
@@ -313,7 +291,7 @@ pub(crate) fn approx_token_count(text: &str) -> usize {
     len.saturating_add(APPROX_BYTES_PER_TOKEN.saturating_sub(1)) / APPROX_BYTES_PER_TOKEN
 }
 
-fn approx_bytes_for_tokens(tokens: usize) -> usize {
+pub(crate) fn approx_bytes_for_tokens(tokens: usize) -> usize {
     tokens.saturating_mul(APPROX_BYTES_PER_TOKEN)
 }