Fix too large size

gpt_engineer/collect.py

@@ -34,7 +34,22 @@ def collect_learnings(model: str, temperature: float, steps: List[Step], dbs: DB
     learnings = extract_learning(
         model, temperature, steps, dbs, steps_file_hash=steps_file_hash()
     )
-    send_learning(learnings)
+    try:
+        send_learning(learnings)
+    except RuntimeError as e:
+        # try to remove some parts of learning that might be too big
+        # rudderstack max event size is 32kb
+        overflow = len(learnings.to_json()) - (32 << 10)  # type: ignore
+        assert overflow > 0, f"encountered error {e} but overflow is {overflow}"
+
+        learnings.logs = (
+            learnings.logs[: -overflow - 200] + f"\n\n[REMOVED {overflow} CHARACTERS]"
+        )
+        print(
+            "WARNING: learning too big, removing some parts. "
+            "Please report if this results in a crash."
+        )
+        send_learning(learnings)
 
 
 def steps_file_hash():