fix(otel): stamp gen_ai.input/output.messages on v2 spans (#30548)

The canonical GenAI mapper's _LLM_CALL_ATTRS table had no extractors for
gen_ai.input.messages or gen_ai.output.messages, so V2 LLM spans never carried
prompt or completion content even when capture was enabled via
OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT=span_and_event. The request
and response bodies were already captured onto LLMCallSpanData.messages_in and
choices_out, but the mapper never read them.

Add the two extractors, serializing messages_in and output_messages(d) through
serialize_messages so the keys are omitted when content capture is off and the
spans stay sparse.

Resolves LIT-3788

litellm/integrations/otel/mappers/genai.py

@@ -10,7 +10,12 @@ table: one lambda per mapping operation, applied against the typed span data.
 from typing import Callable
 
 from litellm.integrations.otel.mappers.base import AttributeMap, AttrValue, SpanData
-from litellm.integrations.otel.mappers.utils import collect, drop_none
+from litellm.integrations.otel.mappers.utils import (
+    collect,
+    drop_none,
+    output_messages,
+    serialize_messages,
+)
 from litellm.integrations.otel.model.payloads import (
     GuardrailSpanData,
     LLMCallSpanData,
@@ -47,6 +52,8 @@ class GenAIMapper:
             else None
         ),
         GenAI.REQUEST_SEED: lambda d: d.request_params.seed,
+        GenAI.INPUT_MESSAGES: lambda d: serialize_messages(d.messages_in),
+        GenAI.OUTPUT_MESSAGES: lambda d: serialize_messages(output_messages(d)),
         GenAI.RESPONSE_MODEL: lambda d: d.response_model,
         GenAI.RESPONSE_ID: lambda d: d.response_id,
         GenAI.RESPONSE_FINISH_REASONS: lambda d: (

tests/test_litellm/integrations/otel/test_otel_v2_components.py

@@ -2,6 +2,8 @@
 baggage helpers, metrics, the typed coercion helpers, mapper branches, span-name
 builders, and the registry validator's failure paths. Needs the OTel SDK."""
 
+import json
+
 import pytest
 
 pytest.importorskip("opentelemetry")
@@ -225,6 +227,47 @@ def test_genai_mapper_all_request_params():
     assert attrs["server.port"] == 443
 
 
+def test_genai_mapper_stamps_input_output_messages():
+    data = LLMCallSpanData(
+        operation=GenAIOperation.CHAT,
+        provider="openai",
+        request_model="gpt-4o",
+        response_model="gpt-4o-2024",
+        response_id="resp_1",
+        request_params=LLMRequestParams(),
+        usage=LLMUsage(),
+        finish_reasons=("stop",),
+        error=None,
+        response_cost=None,
+        server=None,
+        identity=RequestIdentity(call_id="c1"),
+        messages_in=(
+            {"role": "system", "content": "Be concise."},
+            {"role": "user", "content": "What's the weather?"},
+        ),
+        choices_out=(
+            {
+                "finish_reason": "stop",
+                "message": {"role": "assistant", "content": "Sunny."},
+            },
+        ),
+    )
+    attrs = GenAIMapper().map(data)
+    assert json.loads(attrs[GenAI.INPUT_MESSAGES]) == [
+        {"role": "system", "content": "Be concise."},
+        {"role": "user", "content": "What's the weather?"},
+    ]
+    assert json.loads(attrs[GenAI.OUTPUT_MESSAGES]) == [
+        {"role": "assistant", "content": "Sunny."}
+    ]
+
+
+def test_genai_mapper_omits_messages_when_content_not_captured():
+    attrs = GenAIMapper().map(_full_llm_call())
+    assert GenAI.INPUT_MESSAGES not in attrs
+    assert GenAI.OUTPUT_MESSAGES not in attrs
+
+
 def test_genai_mapper_cost_breakdown():
     from litellm.integrations.otel.model.semconv import LiteLLM