Merge pull request #6650 from happy5318/fix/preflight-compression-false-positive

fix: stop false Compressing-context card on preflight-compression sessions

api/streaming.py

@@ -566,8 +566,11 @@ def _is_agent_compression_start_status(kind: str, message: str) -> bool:
     auto-compression.
 
     Positive markers below match the agent emitters in hermes-agent
-    (``turn_context`` preflight, ``conversation_loop`` pre-API / 413 / too-large,
-    ``conversation_compression`` compaction status). Explicitly reject skip /
+    (``conversation_loop`` pre-API / 413 / too-large,
+    ``conversation_compression`` compaction status). Preflight compression
+    (``turn_context``) is intentionally excluded — the later authoritative
+    ``Compacting context`` marker is the signal that compression actually
+    proceeded. Explicitly reject skip /
     defer notices so "Skipping preflight compression…" never surfaces as a
     running compress divider.
     """
@@ -587,8 +590,7 @@ def _is_agent_compression_start_status(kind: str, message: str) -> bool:
     if 'compressed' in m and 'compressing' not in m and 'compression attempt' not in m:
         return False
     return (
-        'preflight compression:' in m
-        or 'pre-api compression:' in m
+        'pre-api compression:' in m
         or 'compacting context' in m
         or 'context too large' in m
         or '— compressing (' in m

tests/test_auto_compression_card.py

@@ -473,7 +473,13 @@ def test_agent_status_callback_emits_compressing_and_warning_events():
 
 def test_agent_compression_start_status_matches_real_emitters_only():
     # Real start notices from hermes-agent emitters
-    assert _is_agent_compression_start_status(
+    # Preflight compression is intentionally excluded — the later
+        # authoritative ``Compacting context`` marker from
+        # conversation_compression is the signal that compression
+        # actually proceeded, and the preflight status can fire even
+        # when compression exits before compaction (e.g. durable
+        # guard refresh, Codex-Hermes mode with no active thread).
+    assert not _is_agent_compression_start_status(
         "lifecycle",
         "📦 Preflight compression: ~101,000 tokens >= 96,000 threshold. This may take a moment.",
     )