v6.7.0 proposal

[dd-octo-sts]

Features

• LLM Observability: Add support for OpenAI Agents (trace-processor) #8044

Fixes

• Feature Flags: Allow custom agentless endpoints #9481
• General: Make peer service finalization idempotent #9364
• LLM Observability: Llm observability traces have custom trace IDs #9460
• standalone: Stamp _dd.apm.enabled on every span #9506
• webpack: Opt out of typescript auto resolve #9521

[dd-octo-sts]

Overall package size

Self size: 7.52 MB
Deduped: 8.18 MB
No deduping: 8.18 MB

Dependency sizes

| name | version | self size | total size | |------|---------|-----------|------------| | import-in-the-middle | 3.3.2 | 124.41 kB | 440.65 kB | | opentracing | 0.14.7 | 194.81 kB | 194.81 kB | | dc-polyfill | 0.1.11 | 25.74 kB | 25.74 kB |

_{🤖 This report was automatically generated by heaviest-objects-in-the-universe}

[datadog-official]

More details

The changed runtime paths behaved correctly across focused LLMObs, OpenAI provider, OpenFeature, peer-service, span-processor, and plugin-structure scenarios. The only unavailable check was the versioned OpenAI Agents integration suite because its external fixtures are not materialized in this sandbox.

Was this helpful? React 👍 or 👎

_{📊 Validated against 8 scenarios · Open Bits AI session}

_{🤖 Datadog Autotest · Commit 3adb270 · What is Autotest? · @DataDog review to ask questions · Any feedback? Reach out in #autotest}

[pr-commenter]

Benchmarks

Benchmark execution time: 2026-07-24 17:34:59

Comparing candidate commit 3adb270 in PR branch v6.7.0-proposal with baseline commit 263c930 in branch v6.x.

📊 Benchmarking dashboard

Found 3 performance improvements and 0 performance regressions! Performance is the same for 2311 metrics, 44 unstable metrics.

Explanation

This is an A/B test comparing a candidate commit's performance against that of a baseline commit. Performance changes are noted in the tables below as:

• 🟩 = significantly better candidate vs. baseline
• 🟥 = significantly worse candidate vs. baseline

We compute a confidence interval (CI) over the relative difference of means between metrics from the candidate and baseline commits, considering the baseline as the reference.

If the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD), the change is considered significant.

Feel free to reach out to #apm-benchmarking-platform on Slack if you have any questions.

More details about the CI and significant changes

You can imagine this CI as a range of values that is likely to contain the true difference of means between the candidate and baseline commits.

CIs of the difference of means are often centered around 0%, because often changes are not that big:

---------------------------------(------|---^--------)-------------------------------->
                              -0.6%    0%  0.3%     +1.2%
                                 |          |        |
         lower bound of the CI --'          |        |
sample mean (center of the CI) -------------'        |
         upper bound of the CI ----------------------'

As described above, a change is considered significant if the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD).

For instance, for an execution time metric, this confidence interval indicates a significantly worse performance:

----------------------------------------|---------|---(---------^---------)---------->
                                       0%        1%  1.3%      2.2%      3.1%
                                                  |   |         |         |
       significant impact threshold --------------'   |         |         |
                      lower bound of CI --------------'         |         |
       sample mean (center of the CI) --------------------------'         |
                      upper bound of CI ----------------------------------'

scenario:llmobs-span-processor-agent-26

• 🟩 cpu_user_time [-463.923ms; -172.302ms] or [-11.689%; -4.341%]
• 🟩 execution_time [-479.621ms; -177.072ms] or [-11.655%; -4.303%]
• 🟩 throughput [+9783.960op/s; +26785.224op/s] or [+4.194%; +11.482%]

Unstable benchmarks

These benchmarks have a confidence interval too wide to call a change; treat them as noise rather than signal.

scenario:appsec-appsec-enabled-24

• unstable execution_time [-209.629ms; +213.290ms] or [-7.862%; +7.999%]

scenario:appsec-appsec-enabled-26

• unstable execution_time [-234670.166µs; +235633.233µs] or [-9.239%; +9.277%]

scenario:appsec-appsec-enabled-with-attacks-26

• unstable execution_time [-183304.705µs; +181501.939µs] or [-6.311%; +6.249%]

scenario:appsec-control-20

• unstable execution_time [-113.776ms; +130.843ms] or [-6.978%; +8.024%]

scenario:appsec-control-24

• unstable execution_time [-111947.734µs; +111694.300µs] or [-8.995%; +8.975%]

scenario:appsec-control-26

• unstable execution_time [-120729.315µs; +121973.715µs] or [-9.752%; +9.852%]

scenario:appsec-iast-no-vulnerability-iast-enabled-default-config-20

• unstable execution_time [-11.755ms; +18.900ms] or [-4.486%; +7.212%]

scenario:debugger-line-probe-with-snapshot-default-24

• unstable cpu_user_time [-2026.792ms; +3245.456ms] or [-24.330%; +38.960%]
• unstable execution_time [-2030.290ms; +3252.162ms] or [-22.439%; +35.944%]
• unstable instructions [-17.1G instructions; +27.2G instructions] or [-25.203%; +40.197%]
• unstable max_rss_usage [-8.320MB; +12.985MB] or [-5.297%; +8.268%]
• unstable throughput [-855.624op/s; +531.401op/s] or [-23.490%; +14.589%]

scenario:debugger-line-probe-with-snapshot-default-26

• unstable cpu_user_time [-2636.073ms; +4125.758ms] or [-27.620%; +43.229%]
• unstable execution_time [-2646.291ms; +4137.807ms] or [-25.752%; +40.267%]
• unstable instructions [-23.4G instructions; +37.2G instructions] or [-29.465%; +46.778%]
• unstable max_rss_usage [-9.900MB; +13.194MB] or [-6.262%; +8.346%]
• unstable throughput [-810.557op/s; +526.141op/s] or [-25.131%; +16.312%]

scenario:debugger-line-probe-with-snapshot-minimal-24

• unstable cpu_user_time [-2.767s; +0.364s] or [-31.083%; +4.083%]
• unstable execution_time [-2.778s; +0.381s] or [-28.880%; +3.962%]
• unstable instructions [-23.4G instructions; +3.1G instructions] or [-32.158%; +4.294%]
• unstable throughput [-104.119op/s; +733.584op/s] or [-2.978%; +20.981%]

scenario:debugger-line-probe-with-snapshot-minimal-26

• unstable cpu_user_time [-2661.249ms; +4218.493ms] or [-27.923%; +44.263%]
• unstable execution_time [-2660.275ms; +4247.263ms] or [-25.886%; +41.329%]
• unstable instructions [-23.4G instructions; +37.2G instructions] or [-29.383%; +46.831%]
• unstable max_rss_usage [-8.797MB; +14.704MB] or [-5.555%; +9.285%]
• unstable throughput [-829.765op/s; +518.552op/s] or [-25.699%; +16.060%]

scenario:debugger-line-probe-without-snapshot-24

• unstable cpu_user_time [-2014.569ms; +3175.880ms] or [-24.371%; +38.420%]
• unstable execution_time [-2059.420ms; +3216.087ms] or [-22.924%; +35.799%]
• unstable instructions [-17.2G instructions; +27.2G instructions] or [-25.431%; +40.254%]
• unstable max_rss_usage [-8.246MB; +12.929MB] or [-5.256%; +8.242%]
• unstable throughput [-854.497op/s; +554.161op/s] or [-23.291%; +15.104%]

scenario:debugger-line-probe-without-snapshot-26

• unstable cpu_user_time [-967.681ms; +286.132ms] or [-10.690%; +3.161%]
• unstable execution_time [-956.022ms; +306.026ms] or [-9.789%; +3.133%]
• unstable instructions [-8.4G instructions; +2.6G instructions] or [-11.174%; +3.429%]
• unstable throughput [-84.873op/s; +263.196op/s] or [-2.564%; +7.952%]

scenario:dogstatsd-with-tags-20

• unstable cpu_user_time [-320.929ms; +307.144ms] or [-6.795%; +6.503%]
• unstable execution_time [-308182.995µs; +307815.162µs] or [-6.421%; +6.414%]
• unstable throughput [-109839.365op/s; +108783.427op/s] or [-6.289%; +6.228%]

scenario:plugin-aws-sdk-extract-response-body-24

• unstable cpu_user_time [-72.817ms; +86.829ms] or [-4.603%; +5.489%]

scenario:plugin-graphql-long-with-depth-and-collapse-off-20

• unstable max_rss_usage [-37.321MB; +31.505MB] or [-9.291%; +7.843%]

scenario:plugin-graphql-long-with-depth-off-20

• unstable max_rss_usage [-2.896MB; +13.503MB] or [-2.204%; +10.278%]

scenario:plugin-graphql-long-with-depth-off-26

• unstable max_rss_usage [-23295.061KB; +21908.776KB] or [-10.712%; +10.074%]

scenario:plugin-graphql-long-with-depth-on-max-20

• unstable throughput [-3.434op/s; +3.390op/s] or [-5.041%; +4.977%]

scenario:test-optimization-large-suite-20

• unstable max_rss_usage [-4.346MB; +7.638MB] or [-5.353%; +9.409%]

[datadog-official]

🎉 All green!

🧪 All tests passed
❄️ No new flaky tests detected

🎯 Code Coverage (details)
• Patch Coverage: 99.46%
• Overall Coverage: 98.44%

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: 3adb270 | Docs | Datadog PR Page | Give us feedback!}

[codecov]

Codecov Report

❌ Patch coverage is 99.45652% with 6 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (v6.x@263c930). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...ages/datadog-instrumentations/src/openai-agents.js	94.49%	6 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             v6.x    #9523   +/-   ##
=======================================
  Coverage        ?   98.44%           
=======================================
  Files           ?      947           
  Lines           ?   128035           
  Branches        ?    10722           
=======================================
  Hits            ?   126046           
  Misses          ?     1989           
  Partials        ?        0           

Flag	Coverage Δ
aiguard	57.49% <48.31%> (?)
aiguard-integration	56.33% <50.87%> (?)
apm-bucket-0	58.76% <52.80%> (?)
apm-bucket-1	64.02% <52.80%> (?)
apm-bucket-2	62.79% <52.80%> (?)
apm-bucket-3	60.23% <52.80%> (?)
apm-capabilities-tracing	64.30% <34.60%> (?)
apm-integrations-aerospike	56.82% <52.80%> (?)
apm-integrations-confluentinc-kafka-javascript	61.67% <52.80%> (?)
apm-integrations-couchbase	57.18% <48.31%> (?)
apm-integrations-http	62.80% <52.80%> (?)
apm-integrations-kafkajs	62.28% <52.80%> (?)
apm-integrations-next	59.27% <48.31%> (?)
apm-integrations-prisma	58.75% <52.80%> (?)
appsec	72.93% <52.80%> (?)
appsec-express_fastify_graphql	70.41% <52.80%> (?)
appsec-integration	51.50% <45.95%> (?)
appsec-kafka_ldapjs_lodash	64.00% <52.80%> (?)
appsec-mongodb-core_mongoose_mysql	67.72% <52.80%> (?)
appsec-next	57.60% <52.80%> (?)
appsec-node-serialize_passport_postgres	67.39% <52.80%> (?)
appsec-sourcing_stripe_template	65.74% <52.80%> (?)
debugger	64.92% <52.80%> (?)
instrumentations-bucket-0	51.83% <47.19%> (?)
instrumentations-bucket-1	60.32% <48.31%> (?)
instrumentations-bucket-10	62.12% <52.80%> (?)
instrumentations-bucket-11	51.75% <47.19%> (?)
instrumentations-bucket-12	52.31% <47.19%> (?)
instrumentations-bucket-13	51.86% <47.19%> (?)
instrumentations-bucket-2	53.74% <47.19%> (?)
instrumentations-bucket-3	59.32% <48.31%> (?)
instrumentations-bucket-4	52.40% <47.19%> (?)
instrumentations-bucket-5	57.64% <52.80%> (?)
instrumentations-bucket-6	60.85% <52.80%> (?)
instrumentations-bucket-7	58.52% <52.80%> (?)
instrumentations-bucket-8	59.66% <52.80%> (?)
instrumentations-bucket-9	61.64% <48.31%> (?)
instrumentations-instrumentation-couchbase	51.14% <47.19%> (?)
instrumentations-integration-esbuild	34.27% <43.58%> (?)
llmobs-ai_anthropic_bedrock	62.90% <84.76%> (?)
llmobs-bucket-1	61.71% <80.00%> (?)
llmobs-openai	62.44% <48.99%> (?)
llmobs-openai-agents_vertex-ai	59.79% <97.53%> (?)
llmobs-sdk	65.60% <98.82%> (?)
openfeature	56.18% <50.51%> (?)
openfeature-unit	53.48% <51.54%> (?)
platform-core_esbuild_instrumentations-misc	40.64% <45.45%> (?)
platform-integration	61.50% <48.31%> (?)
platform-shimmer_unit-guardrails_webpack	39.13% <45.45%> (?)
plugins-bucket-0	57.15% <52.80%> (?)
plugins-bucket-1	54.48% <48.31%> (?)
plugins-bucket-11	62.36% <48.31%> (?)
plugins-bucket-18	62.06% <52.80%> (?)
plugins-bucket-19	60.15% <48.31%> (?)
plugins-bucket-20	62.15% <48.31%> (?)
plugins-bucket-4	58.68% <48.31%> (?)
plugins-bullmq_cassandra_cookie	61.85% <52.80%> (?)
plugins-cookie-parser_crypto_dd-trace-api	56.83% <48.31%> (?)
plugins-fetch_fs_generic-pool	58.88% <52.80%> (?)
plugins-google-cloud-pubsub_grpc_handlebars	64.78% <52.80%> (?)
plugins-hapi_hono_ioredis	60.32% <52.80%> (?)
plugins-jest_knex_langgraph	55.70% <48.31%> (?)
plugins-ldapjs_light-my-request_limitd-client	58.61% <52.80%> (?)
plugins-lodash_mariadb_memcached	58.19% <52.80%> (?)
plugins-moleculer_mongodb_mongodb-core	62.01% <52.80%> (?)
plugins-mongoose_multer_mysql	59.18% <52.80%> (?)
plugins-mysql2_nats_node-serialize	60.81% <52.80%> (?)
plugins-opensearch_passport-http_pino	59.65% <52.80%> (?)
plugins-postgres_process_pug	58.38% <52.80%> (?)
plugins-redis_router_sequelize	62.20% <52.80%> (?)
plugins-test-and-upstream-rhea_undici_url	61.73% <52.80%> (?)
plugins-valkey_vm_winston	58.12% <52.80%> (?)
plugins-ws	59.70% <48.31%> (?)
profiling	62.24% <52.80%> (?)
serverless-aws-sdk-aws-sdk	55.15% <52.80%> (?)
serverless-aws-sdk-bedrockruntime	54.86% <52.80%> (?)
serverless-aws-sdk-client	56.53% <52.80%> (?)
serverless-aws-sdk-dynamodb	55.76% <52.80%> (?)
serverless-aws-sdk-eventbridge	49.48% <47.19%> (?)
serverless-aws-sdk-kinesis	59.44% <52.80%> (?)
serverless-aws-sdk-lambda	57.49% <52.80%> (?)
serverless-aws-sdk-s3	55.86% <52.80%> (?)
serverless-aws-sdk-serverless-peer-service	59.84% <52.80%> (?)
serverless-aws-sdk-sns	60.30% <52.80%> (?)
serverless-aws-sdk-sqs	60.72% <52.80%> (?)
serverless-aws-sdk-stepfunctions	55.69% <52.80%> (?)
serverless-aws-sdk-util	51.59% <47.19%> (?)
serverless-bucket-0	54.34% <48.31%> (?)
serverless-bucket-1	59.35% <48.31%> (?)
test-optimization-cucumber	71.98% <48.31%> (?)
test-optimization-cypress	65.87% <48.31%> (?)
test-optimization-jest	73.29% <48.31%> (?)
test-optimization-mocha	73.54% <48.31%> (?)
test-optimization-playwright-playwright-atr	60.46% <48.31%> (?)
test-optimization-playwright-playwright-efd	60.65% <48.31%> (?)
test-optimization-playwright-playwright-final-status	60.62% <48.31%> (?)
test-optimization-playwright-playwright-impacted-tests	60.35% <48.31%> (?)
test-optimization-playwright-playwright-reporting	61.64% <48.31%> (?)
test-optimization-playwright-playwright-test-management	61.22% <48.31%> (?)
test-optimization-playwright-playwright-test-span	60.52% <48.31%> (?)
test-optimization-selenium	60.13% <48.31%> (?)
test-optimization-testopt	58.59% <48.31%> (?)
test-optimization-vitest	70.33% <48.31%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3adb2703d3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Parent agent workflows to the active span

When an OpenAI Agents run is started inside an already traced request or job, this direct core-tracer startSpan() call does not inherit the active legacy scope, so the workflow span (and all of its agent/LLM/tool children) becomes a separate root trace instead of staying under the surrounding application span. Pass the active span as childOf for the workflow root, or create it through the plugin tracing helper that applies the active parent.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Suppress duplicate OpenAI LLMObs model spans

When LLMObs is enabled for the normal OpenAIResponsesModel path, the underlying openai client still goes through the OpenAI plugin's responses.create LLMObs registration while this code also registers the agents-core response span as an LLM span. One model invocation therefore emits two LLMObs model-call events (and the lower-level one is not parented through the agents LLMObs hierarchy), which can double-count requests/tokens; add a suppression flag around the agent-owned model call or otherwise skip the OpenAI LLMObs span while keeping its APM span.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Stamp spans with the openai-agents integration

These direct core-tracer spans only set the component tag; because no integrationName is passed, Span.finish() stamps _dd.integration as the default opentracing instead of openai-agents. In traced agent runs this breaks per-integration attribution/telemetry for every agent, tool, and LLM span, so pass integrationName: COMPONENT (or route span creation through the plugin tracing helper) for these spans as well as the workflow root.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Add the plugin to APM integration CI

This adds openai-agents to the supported integration list, but .github/workflows/apm-integrations.yml has no matching openai-agents/@openai/agents job or PLUGINS entry, so the new package tests will not be exercised by the APM integration matrix. Please register the new plugin in that workflow alongside the other plugin registration points.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Handle Chat Completions model spans

The supported Agents SDK also exposes OpenAIChatCompletionsModel, and that path emits agents-core spans with spanData.type === 'generation' instead of response; because this map has no generation entry, onSpanStart() drops those model calls before creating any dd-trace/LLMObs model span. Agents configured for Chat Completions will therefore miss token/model events, so map generation to the LLM kind and tag its input/output/model fields like the response path.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Keep the agents provider paired per module copy

In a process with two installed copies of @openai/agents, this single module-level reference is overwritten by whichever copy loads last, while every patched @openai/agents-openai model wrapper reads it later. A call from the earlier copy then asks the wrong trace provider for getCurrentSpan(), so agentsCoreSpanId is missing or from another run and the OpenAI request span is no longer bound under the correct agent span; keep the provider/module reference scoped to the patched model copy instead of one global.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Create the LLMObs tagger when LLMObs is enabled

If the tracer starts with LLMObs disabled and the app later uses the still-supported llmobs.enable() API after this plugin has been constructed, the shared config flips DD_LLMOBS_ENABLED to true but #tagger stays undefined forever because it was only initialized here. In that scenario openai-agents continues to emit APM spans but never registers or tags LLMObs events, unlike the other LLMObs plugins that keep a tagger and let it read the updated config; instantiate the tagger unconditionally or when enabling.

Useful? React with 👍 / 👎.