fix(llmobs): replace trace processor with event listener

[Kyle-Verhoog]

The LLMObs service formerly depended on the TraceProcessor interface in the tracer. This was problematic due to sharing a dependency with the public API. As such, users could configure a trace filter (under the hood is a trace processor) and overwrite the LLMObs TraceProcessor.

Instead, the tracer can emit span start and finish events which the LLMObs service listens to and acts on, as proposed here.

The gotcha is that the LLMObs service no longer has a way to drop traces when run in agentless mode, which only LLMObs supports. Instead, we encourage users to explicitly turn off APM which carries the benefit of clarity since this was implicit before.

Checklist

• 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
• The change includes or references documentation updates if necessary
• Backport labels are set (if applicable)

Reviewer Checklist

• 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 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

[github-actions]

CODEOWNERS have been resolved as:

.riot/requirements/16562eb.txt                                          @DataDog/apm-python
releasenotes/notes/fix-llmobs-processor-4afd715a84323d32.yaml           @DataDog/apm-python
ddtrace/_trace/tracer.py                                                @DataDog/apm-sdk-api-python
ddtrace/llmobs/_llmobs.py                                               @DataDog/ml-observability
ddtrace/llmobs/_utils.py                                                @DataDog/ml-observability
riotfile.py                                                             @DataDog/apm-python
tests/llmobs/conftest.py                                                @DataDog/ml-observability
tests/llmobs/test_llmobs.py                                             @DataDog/ml-observability
tests/llmobs/test_llmobs_ragas_faithfulness_evaluator.py                @DataDog/ml-observability
tests/llmobs/test_llmobs_service.py                                     @DataDog/ml-observability
tests/llmobs/test_llmobs_span_agent_writer.py                           @DataDog/ml-observability
tests/llmobs/test_llmobs_span_agentless_writer.py                       @DataDog/ml-observability
tests/llmobs/test_propagation.py                                        @DataDog/ml-observability

[Kyle-Verhoog]

the existing tests for the trace processor are coupled to the implementation so i'm gonna open up a refactor on those first instead of trying to migrate them at the same time increasing the risk of a regression

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-01-08 23:44:58

Comparing candidate commit 047e147 in PR branch llmobs/span-processor with baseline commit 6c61f40 in branch main.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 148 metrics, 2 unstable metrics.

[Kyle-Verhoog]

done in #11784

[datadog-dd-trace-py-rkomorn]

Datadog Report

Branch report: llmobs/span-processor
Commit report: 047e147
Test service: dd-trace-py

✅ 0 Failed, 1555 Passed, 0 Skipped, 24m 49.5s Total duration (13m 53.33s time saved)

[Yun-Kim]

Instead, we encourage users to explicitly turn off APM which carries the benefit of clarity since this was implicit before.

@Kyle-Verhoog should we add a release note specifying that DD_TRACE_ENABLED should be set to false for agentless users? Or would you rather we do this implicitly for agentless customers?

[Kyle-Verhoog]

yes let's add a release note for the change

[mabdinur]

Left some questions but overall looks good to me