fix(sdk): keep oversized entity attributes exportable

Author: Whxuan0701

packages/traceloop-sdk/README.md

@@ -2,6 +2,11 @@
 
 Traceloop’s Python SDK allows you to easily start monitoring and debugging your LLM execution. Tracing is done in a non-intrusive way, built on top of OpenTelemetry. You can choose to export the traces to Traceloop, or to your existing observability stack.
 
+Entity input and output attributes are bounded to 1,000,000 characters by
+default so oversized payloads remain visible to OTLP backends instead of being
+dropped as invalid attributes. Set `OTEL_SPAN_ATTRIBUTE_VALUE_LENGTH_LIMIT` to
+use a different positive limit.
+
 ```python
 Traceloop.init(app_name="joke_generation_service")
 

packages/traceloop-sdk/tests/test_tasks.py

@@ -206,6 +206,22 @@ def no_truncation_task(long_input):
     assert output_data == result
 
 
+def test_json_large_content_is_bounded_without_otel_limit(exporter, monkeypatch):
+    """Large entity payloads remain exportable when no OTel limit is configured."""
+    monkeypatch.delenv("OTEL_SPAN_ATTRIBUTE_VALUE_LENGTH_LIMIT", raising=False)
+
+    @task(name="large_content_task")
+    def large_content_task():
+        return "x" * 1_100_000
+
+    large_content_task()
+
+    span = exporter.get_finished_spans()[0]
+    output_json = span.attributes[SpanAttributes.TRACELOOP_ENTITY_OUTPUT]
+    assert len(output_json) == 1_000_000
+    assert output_json.startswith('"' + ("x" * 999_999))
+
+
 def test_json_truncation_with_invalid_otel_limit(exporter, monkeypatch):
     """Test that invalid OTEL_SPAN_ATTRIBUTE_VALUE_LENGTH_LIMIT values are ignored"""
     # Set environment variable to invalid value

packages/traceloop-sdk/traceloop/sdk/decorators/base.py

@@ -31,20 +31,27 @@
 
 F = TypeVar("F", bound=Callable[..., Any])
 
+# Keep entity payloads below the size at which common OTLP backends reject the
+# entire attribute. Users can opt into the standard OTel limit with the env var.
+_DEFAULT_SPAN_ATTRIBUTE_VALUE_LENGTH_LIMIT = 1_000_000
+
 
 def _truncate_json_if_needed(json_str: str) -> str:
     """
     Truncate JSON string if it exceeds OTEL_SPAN_ATTRIBUTE_VALUE_LENGTH_LIMIT;
     truncation may yield an invalid JSON string, which is expected for logging purposes.
     """
     limit_str = os.getenv("OTEL_SPAN_ATTRIBUTE_VALUE_LENGTH_LIMIT")
+    limit = _DEFAULT_SPAN_ATTRIBUTE_VALUE_LENGTH_LIMIT
     if limit_str:
         try:
-            limit = int(limit_str)
-            if limit > 0 and len(json_str) > limit:
-                return json_str[:limit]
+            configured_limit = int(limit_str)
+            if configured_limit > 0:
+                limit = configured_limit
         except ValueError:
             pass
+    if len(json_str) > limit:
+        return json_str[:limit]
     return json_str