openai/codex
1
0
Fork 0
mirror of https://github.com/openai/codex.git synced 2026-10-30 19:18:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: max token usage with estimate for compaction

Browse Source
This commit is contained in:
jif-oai
2025-12-19 12:50:30 +01:00
parent 6f94a90797
commit 90a9597afa
1 changed files with 16 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
codex-rs/core/src/codex.rs
+16 -2
View File
@@ -764,6 +764,16 @@ impl Session {
state.get_total_token_usage()
}
async fn get_total_token_usage_with_estimate(&self, turn_context: &TurnContext) -> i64 {
let reported = self.get_total_token_usage().await;
let estimated = self
.clone_history()
.await
.estimate_token_count(turn_context)
.unwrap_or(0);
reported.max(estimated)
}
async fn record_initial_history(&self, conversation_history: InitialHistory) {
let turn_context = self.new_default_turn().await;
match conversation_history {
@@ -2233,7 +2243,9 @@ pub(crate) async fn run_task(
.get_model_family()
.auto_compact_token_limit()
.unwrap_or(i64::MAX);
let total_usage_tokens = sess.get_total_token_usage().await;
let total_usage_tokens = sess
.get_total_token_usage_with_estimate(&turn_context)
.await;
if total_usage_tokens >= auto_compact_limit {
run_auto_compact(&sess, &turn_context).await;
}
@@ -2315,7 +2327,9 @@ pub(crate) async fn run_task(
needs_follow_up,
last_agent_message: turn_last_agent_message,
} = turn_output;
let total_usage_tokens = sess.get_total_token_usage().await;
let total_usage_tokens = sess
.get_total_token_usage_with_estimate(&turn_context)
.await;
let token_limit_reached = total_usage_tokens >= auto_compact_limit;
// as long as compaction works well in getting us way below the token limit, we shouldn't worry about being in an infinite loop.