-
Notifications
You must be signed in to change notification settings - Fork 468
fix(llmobs): safely output format bedrock cohere rerank spans [backport 3.17] #15137
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
## Description Closes #14575. <!-- Provide an overview of the change and motivation for the change --> Adds safely accessing output response messages in the bedrock integration. This happens when cohere rerank models are invoked, since cohere rerank responses lack a response["text"] field and will return an empty list. ## Testing Added a test invoking cohere rerank models. <!-- Describe your testing strategy or note what tests are included --> ## Risks <!-- Note any risks associated with this change, or "None" if no risks --> ## Additional Notes <!-- Any other information that would be helpful for reviewers --> (cherry picked from commit 191f10e) Signed-off-by: Yun Kim <yun.kim@datadoghq.com>
|
|
Bootstrap import analysisComparison of import times between this PR and base. SummaryThe average import time from this PR is: 244 ± 3 ms. The average import time from base is: 245 ± 3 ms. The import time difference between this PR and base is: -0.2 ± 0.1 ms. The difference is not statistically significant (z = -1.73). Import time breakdownThe following import paths have shrunk:
|
45ae804 to
41f5ec5
Compare
Performance SLOsComparing candidate backport-15124-to-3.17 (f5f09e1) with baseline 3.17 (cf7b327) 🟡 Near SLO Breach (4 suites)🟡 djangosimple - 30/30✅ appsecTime: ✅ 20.480ms (SLO: <22.300ms -8.2%) vs baseline: ~same Memory: ✅ 65.274MB (SLO: <67.000MB -2.6%) vs baseline: +4.9% ✅ exception-replay-enabledTime: ✅ 1.342ms (SLO: <1.450ms -7.4%) vs baseline: -0.7% Memory: ✅ 64.325MB (SLO: <67.000MB -4.0%) vs baseline: +4.7% ✅ iastTime: ✅ 20.490ms (SLO: <22.250ms -7.9%) vs baseline: ~same Memory: ✅ 65.136MB (SLO: <67.000MB -2.8%) vs baseline: +4.5% ✅ profilerTime: ✅ 15.265ms (SLO: <16.550ms -7.8%) vs baseline: ~same Memory: ✅ 53.752MB (SLO: <54.500MB 🟡 -1.4%) vs baseline: +5.0% ✅ resource-renamingTime: ✅ 20.524ms (SLO: <21.750ms -5.6%) vs baseline: +0.4% Memory: ✅ 65.255MB (SLO: <67.000MB -2.6%) vs baseline: +4.7% ✅ span-code-originTime: ✅ 25.376ms (SLO: <28.200ms 📉 -10.0%) vs baseline: -0.1% Memory: ✅ 67.497MB (SLO: <69.500MB -2.9%) vs baseline: +5.0% ✅ tracerTime: ✅ 20.497ms (SLO: <21.750ms -5.8%) vs baseline: ~same Memory: ✅ 65.195MB (SLO: <67.000MB -2.7%) vs baseline: +4.7% ✅ tracer-and-profilerTime: ✅ 22.164ms (SLO: <23.500ms -5.7%) vs baseline: +0.6% Memory: ✅ 66.506MB (SLO: <67.500MB 🟡 -1.5%) vs baseline: +4.7% ✅ tracer-dont-create-db-spansTime: ✅ 19.332ms (SLO: <21.500ms 📉 -10.1%) vs baseline: -0.3% Memory: ✅ 65.226MB (SLO: <66.000MB 🟡 -1.2%) vs baseline: +4.6% ✅ tracer-minimalTime: ✅ 16.648ms (SLO: <17.500ms -4.9%) vs baseline: +0.2% Memory: ✅ 65.195MB (SLO: <66.000MB 🟡 -1.2%) vs baseline: +4.7% ✅ tracer-nativeTime: ✅ 20.408ms (SLO: <21.750ms -6.2%) vs baseline: +0.1% Memory: ✅ 71.520MB (SLO: <72.500MB 🟡 -1.4%) vs baseline: +4.9% ✅ tracer-no-cachesTime: ✅ 18.530ms (SLO: <19.650ms -5.7%) vs baseline: +0.4% Memory: ✅ 65.251MB (SLO: <67.000MB -2.6%) vs baseline: +4.9% ✅ tracer-no-databasesTime: ✅ 18.814ms (SLO: <20.100ms -6.4%) vs baseline: ~same Memory: ✅ 65.107MB (SLO: <67.000MB -2.8%) vs baseline: +4.6% ✅ tracer-no-middlewareTime: ✅ 20.202ms (SLO: <21.500ms -6.0%) vs baseline: ~same Memory: ✅ 65.169MB (SLO: <67.000MB -2.7%) vs baseline: +4.7% ✅ tracer-no-templatesTime: ✅ 20.262ms (SLO: <22.000ms -7.9%) vs baseline: -0.1% Memory: ✅ 65.284MB (SLO: <67.000MB -2.6%) vs baseline: +4.9% 🟡 errortrackingdjangosimple - 6/6✅ errortracking-enabled-allTime: ✅ 18.073ms (SLO: <19.850ms -9.0%) vs baseline: ~same Memory: ✅ 65.215MB (SLO: <66.500MB 🟡 -1.9%) vs baseline: +4.8% ✅ errortracking-enabled-userTime: ✅ 18.198ms (SLO: <19.400ms -6.2%) vs baseline: +1.1% Memory: ✅ 65.110MB (SLO: <66.500MB -2.1%) vs baseline: +4.8% ✅ tracer-enabledTime: ✅ 18.110ms (SLO: <19.450ms -6.9%) vs baseline: +0.2% Memory: ✅ 65.147MB (SLO: <66.500MB -2.0%) vs baseline: +4.8% 🟡 flasksimple - 18/18✅ appsec-getTime: ✅ 4.603ms (SLO: <4.750ms -3.1%) vs baseline: ~same Memory: ✅ 61.774MB (SLO: <65.000MB -5.0%) vs baseline: +4.9% ✅ appsec-postTime: ✅ 6.606ms (SLO: <6.750ms -2.1%) vs baseline: -0.1% Memory: ✅ 61.813MB (SLO: <65.000MB -4.9%) vs baseline: +5.0% ✅ appsec-telemetryTime: ✅ 4.588ms (SLO: <4.750ms -3.4%) vs baseline: ~same Memory: ✅ 61.814MB (SLO: <65.000MB -4.9%) vs baseline: +4.9% ✅ debuggerTime: ✅ 1.858ms (SLO: <2.000ms -7.1%) vs baseline: -0.3% Memory: ✅ 45.259MB (SLO: <47.000MB -3.7%) vs baseline: +4.3% ✅ iast-getTime: ✅ 1.865ms (SLO: <2.000ms -6.7%) vs baseline: -0.4% Memory: ✅ 42.448MB (SLO: <49.000MB 📉 -13.4%) vs baseline: +5.1% ✅ profilerTime: ✅ 1.914ms (SLO: <2.100ms -8.9%) vs baseline: -0.2% Memory: ✅ 46.510MB (SLO: <47.000MB 🟡 -1.0%) vs baseline: +4.6% ✅ resource-renamingTime: ✅ 3.373ms (SLO: <3.650ms -7.6%) vs baseline: -0.2% Memory: ✅ 52.081MB (SLO: <53.500MB -2.7%) vs baseline: +4.7% ✅ tracerTime: ✅ 3.354ms (SLO: <3.650ms -8.1%) vs baseline: -0.3% Memory: ✅ 52.022MB (SLO: <53.500MB -2.8%) vs baseline: +4.4% ✅ tracer-nativeTime: ✅ 3.356ms (SLO: <3.650ms -8.1%) vs baseline: -0.4% Memory: ✅ 58.209MB (SLO: <60.000MB -3.0%) vs baseline: +5.0% 🟡 otelsdkspan - 24/24✅ add-eventTime: ✅ 41.207ms (SLO: <42.000ms 🟡 -1.9%) vs baseline: +1.7% Memory: ✅ 34.780MB (SLO: <39.000MB 📉 -10.8%) vs baseline: +5.0% ✅ add-linkTime: ✅ 36.557ms (SLO: <38.550ms -5.2%) vs baseline: +0.3% Memory: ✅ 34.721MB (SLO: <39.000MB 📉 -11.0%) vs baseline: +4.9% ✅ add-metricsTime: ✅ 219.381ms (SLO: <232.000ms -5.4%) vs baseline: +0.1% Memory: ✅ 34.662MB (SLO: <39.000MB 📉 -11.1%) vs baseline: +4.6% ✅ add-tagsTime: ✅ 211.857ms (SLO: <221.600ms -4.4%) vs baseline: -0.4% Memory: ✅ 34.623MB (SLO: <39.000MB 📉 -11.2%) vs baseline: +4.1% ✅ get-contextTime: ✅ 29.882ms (SLO: <31.300ms -4.5%) vs baseline: +2.7% Memory: ✅ 34.603MB (SLO: <39.000MB 📉 -11.3%) vs baseline: +4.1% ✅ is-recordingTime: ✅ 29.897ms (SLO: <31.000ms -3.6%) vs baseline: +2.2% Memory: ✅ 34.524MB (SLO: <39.000MB 📉 -11.5%) vs baseline: +4.0% ✅ record-exceptionTime: ✅ 63.290ms (SLO: <65.850ms -3.9%) vs baseline: ~same Memory: ✅ 34.760MB (SLO: <39.000MB 📉 -10.9%) vs baseline: +4.6% ✅ set-statusTime: ✅ 32.515ms (SLO: <34.150ms -4.8%) vs baseline: +1.1% Memory: ✅ 34.662MB (SLO: <39.000MB 📉 -11.1%) vs baseline: +4.3% ✅ startTime: ✅ 28.788ms (SLO: <30.150ms -4.5%) vs baseline: -0.6% Memory: ✅ 34.465MB (SLO: <39.000MB 📉 -11.6%) vs baseline: +3.8% ✅ start-finishTime: ✅ 34.153ms (SLO: <35.350ms -3.4%) vs baseline: +0.1% Memory: ✅ 34.642MB (SLO: <39.000MB 📉 -11.2%) vs baseline: +4.1% ✅ start-finish-telemetryTime: ✅ 34.084ms (SLO: <35.450ms -3.9%) vs baseline: -0.6% Memory: ✅ 34.760MB (SLO: <39.000MB 📉 -10.9%) vs baseline: +4.6% ✅ update-nameTime: ✅ 31.919ms (SLO: <33.400ms -4.4%) vs baseline: +2.1% Memory: ✅ 34.603MB (SLO: <39.000MB 📉 -11.3%) vs baseline: +4.1%
|
Backport 191f10e from #15124 to 3.17.
Description
Closes #14575.
Adds safely accessing output response messages in the bedrock integration. This happens when cohere rerank models are invoked, since cohere rerank responses lack a response["text"] field and will return an empty list.
Testing
Added a test invoking cohere rerank models.
Risks
Additional Notes