feat(llmobs): support sending 5mb events (#14172)

Support sending 5mb events in the python sdk

### Testing

All 5mb events show up from this script
```
for i in range(100):
    with LLMObs.agent(
        "test"
    ):
        LLMObs.annotate(
            input_data = "heeee" * 970_000,
            output_data = "bye",
        )
```
[We get all 200 events
](https://dd.datad0g.com/llm/traces?query=%40ml_app%3A5mb-load-test-sdk%20%40event_type%3Aspan%20%40parent_id%3Aundefined&agg_m=count&agg_m_source=base&agg_t=count&fromUser=false&llmtgc=categories&llmtgl=volume&spanId=199578092099801838&start=1753304058676&end=1753908858676&paused=false)

<img width="1123" height="256" alt="image"
src="https://github.com/user-attachments/assets/c91ad30f-c979-431a-93aa-81fa10e7df75"
/>


## Checklist
- [x] PR author has checked that all the criteria below are met
- The PR description includes an overview of the change
- The PR description articulates the motivation for the change
- The change includes tests OR the PR description describes a testing
strategy
- The PR description notes risks associated with the change, if any
- Newly-added code is easy to change
- The change follows the [library release note
guidelines](https://ddtrace.readthedocs.io/en/stable/releasenotes.html)
- The change includes or references documentation updates if necessary
- Backport labels are set (if
[applicable](https://ddtrace.readthedocs.io/en/latest/contributing.html#backporting))

## Reviewer Checklist
- [x] Reviewer has checked that all the criteria below are met 
- Title is accurate
- All changes are related to the pull request's stated goal
- Avoids breaking
[API](https://ddtrace.readthedocs.io/en/stable/versioning.html#interfaces)
changes
- Testing strategy adequately addresses listed risks
- Newly-added code is easy to change
- Release note makes sense to a user of the library
- If necessary, author has acknowledged and discussed the performance
implications of this PR as reported in the benchmarks PR comment
- Backport labels are set in a manner that is consistent with the
[release branch maintenance
policy](https://ddtrace.readthedocs.io/en/latest/contributing.html#backporting)

Author: lievan

ddtrace/llmobs/_constants.py

@@ -59,7 +59,7 @@
 DD_SITES_NEEDING_APP_SUBDOMAIN = {"datadoghq.com", "datadoghq.eu", "ddog-gov.com"}
 
 EVP_PAYLOAD_SIZE_LIMIT = 5 << 20  # 5MB (actual limit is 5.1MB)
-EVP_EVENT_SIZE_LIMIT = (1 << 20) - 1024  # 999KB (actual limit is 1MB)
+EVP_EVENT_SIZE_LIMIT = 5_000_000  # 5MB LLM Obs event size limit
 
 EXPERIMENT_CSV_FIELD_MAX_SIZE = 10 * 1024 * 1024
 

ddtrace/llmobs/_writer.py

@@ -523,7 +523,7 @@ def enqueue(self, event: LLMObsSpanEvent) -> None:
         should_truncate = raw_event_size >= EVP_EVENT_SIZE_LIMIT
         if should_truncate:
             logger.warning(
-                "dropping event input/output because its size (%d) exceeds the event size limit (1MB)",
+                "dropping event input/output because its size (%d) exceeds the event size limit (5MB)",
                 raw_event_size,
             )
             event = _truncate_span_event(event)

releasenotes/notes/bump-llmobs-5mb-617c24b81a2db2eg.yaml

@@ -0,0 +1,4 @@
+---
+features:
+  - |
+    LLM Observability: Increases span event size limit from 1MB to 5MB.

tests/llmobs/_utils.py

@@ -446,7 +446,7 @@ def _large_event():
             "output": {
                 "messages": [
                     {
-                        "content": "A" * 900_000,
+                        "content": "A" * 2_000_000,
                         "role": "assistant",
                     },
                 ]
@@ -477,14 +477,14 @@ def _oversized_llm_event():
                         "role": "system",
                         "content": "You are an evil dark lord looking for his one ring to rule them all",
                     },
-                    {"role": "user", "content": "A" * 700_000},
+                    {"role": "user", "content": "A" * 2_600_000},
                 ],
                 "parameters": {"temperature": 0.9, "max_tokens": 256},
             },
             "output": {
                 "messages": [
                     {
-                        "content": "A" * 700_000,
+                        "content": "A" * 2_600_000,
                         "role": "assistant",
                     },
                 ]
@@ -507,8 +507,8 @@ def _oversized_workflow_event():
         "status": "ok",
         "meta": {
             "span.kind": "workflow",
-            "input": {"value": "A" * 700_000},
-            "output": {"value": "A" * 700_000},
+            "input": {"value": "A" * 2_600_000},
+            "output": {"value": "A" * 2_600_000},
         },
         "metrics": {"input_tokens": 64, "output_tokens": 128, "total_tokens": 192},
     }
@@ -527,8 +527,8 @@ def _oversized_retrieval_event():
         "status": "ok",
         "meta": {
             "span.kind": "retrieval",
-            "input": {"documents": {"content": "A" * 700_000}},
-            "output": {"value": "A" * 700_000},
+            "input": {"documents": {"content": "A" * 2_600_000}},
+            "output": {"value": "A" * 2_600_000},
         },
         "metrics": {"input_tokens": 64, "output_tokens": 128, "total_tokens": 192},
     }

tests/llmobs/test_llmobs_span_agent_writer.py

@@ -49,14 +49,11 @@ def test_flush_queue_when_event_cause_queue_to_exceed_payload_limit(mock_send_pa
     llmobs_span_writer.enqueue(_large_event())
     llmobs_span_writer.enqueue(_large_event())
     llmobs_span_writer.enqueue(_large_event())
-    llmobs_span_writer.enqueue(_large_event())
-    llmobs_span_writer.enqueue(_large_event())
-    llmobs_span_writer.enqueue(_large_event())
     llmobs_span_writer.periodic()
     mock_writer_logs.debug.assert_has_calls(
         [
             mock.call("manually flushing buffer because queueing next event will exceed EVP payload limit"),
-            mock.call("encoded %d LLMObs %s events to be sent", 5, "span"),
+            mock.call("encoded %d LLMObs %s events to be sent", 2, "span"),
             mock.call("encoded %d LLMObs %s events to be sent", 1, "span"),
         ],
         any_order=True,
@@ -71,9 +68,9 @@ def test_truncating_oversized_events(mock_send_payload, mock_writer_logs):
     llmobs_span_writer.enqueue(_oversized_workflow_event())
     mock_writer_logs.warning.assert_has_calls(
         [
-            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (1MB)", 1400724),
-            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (1MB)", 1400464),
-            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (1MB)", 1400445),
+            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (5MB)", 5200724),
+            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (5MB)", 5200464),
+            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (5MB)", 5200445),
         ]
     )
 

tests/llmobs/test_llmobs_span_agentless_writer.py

@@ -43,14 +43,11 @@ def test_flush_queue_when_event_cause_queue_to_exceed_payload_limit(mock_send_pa
     llmobs_span_writer.enqueue(_large_event())
     llmobs_span_writer.enqueue(_large_event())
     llmobs_span_writer.enqueue(_large_event())
-    llmobs_span_writer.enqueue(_large_event())
-    llmobs_span_writer.enqueue(_large_event())
-    llmobs_span_writer.enqueue(_large_event())
     llmobs_span_writer.periodic()
     mock_writer_logs.debug.assert_has_calls(
         [
             mock.call("manually flushing buffer because queueing next event will exceed EVP payload limit"),
-            mock.call("encoded %d LLMObs %s events to be sent", 5, "span"),
+            mock.call("encoded %d LLMObs %s events to be sent", 2, "span"),
             mock.call("encoded %d LLMObs %s events to be sent", 1, "span"),
         ],
         any_order=True,
@@ -64,9 +61,9 @@ def test_truncating_oversized_events(mock_writer_logs):
     llmobs_span_writer.enqueue(_oversized_workflow_event())
     mock_writer_logs.warning.assert_has_calls(
         [
-            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (1MB)", 1400724),
-            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (1MB)", 1400464),
-            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (1MB)", 1400445),
+            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (5MB)", 5200724),
+            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (5MB)", 5200464),
+            mock.call("dropping event input/output because its size (%d) exceeds the event size limit (5MB)", 5200445),
         ]
     )