feat: Update OpenAI graph runner to return AgentGraphRunnerResult with GraphMetrics#155
Merged
jsonbailey merged 5 commits intomainfrom May 4, 2026
Merged
Conversation
Merged
3 tasks
fcbcb18 to
9286d53
Compare
9733a28 to
44501e3
Compare
9286d53 to
72fc13e
Compare
44501e3 to
142e041
Compare
72fc13e to
bde4f09
Compare
142e041 to
fb3c0f6
Compare
bde4f09 to
c376011
Compare
fb3c0f6 to
b3547b0
Compare
c376011 to
7f67e4f
Compare
b3547b0 to
1d4ddb2
Compare
7f67e4f to
a89c6a2
Compare
1d4ddb2 to
6201d09
Compare
a89c6a2 to
c69a9ff
Compare
6201d09 to
ef4216c
Compare
c69a9ff to
14cfa92
Compare
ef4216c to
8ecce16
Compare
14cfa92 to
1ed1a44
Compare
8ecce16 to
09af502
Compare
1ed1a44 to
f016b0d
Compare
09af502 to
43bc879
Compare
5527667 to
8035890
Compare
…h GraphMetrics Remove all direct LaunchDarkly tracker calls from OpenAIAgentGraphRunner. The runner now collects per-node metrics via _NodeMetricsAccumulator (a lightweight accumulator replacing the per-node LDAIConfigTracker) and returns AgentGraphRunnerResult with populated GraphMetrics (path, duration_ms, usage, node_metrics). Graph-level and per-node tracking events are emitted by ManagedAgentGraph from the result via track_graph_metrics_of_async and _track_node_metrics. Integration tests in test_tracking_openai_agents.py are updated to run through the full ManagedAgentGraph pipeline (ManagedAgentGraph.run()) so tracking events are emitted by the managed layer as intended. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
43bc879 to
f6df746
Compare
keelerm84
approved these changes
May 4, 2026
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 49cb245. Configure here.
Merged
jsonbailey
added a commit
that referenced
this pull request
May 6, 2026
🤖 I have created a release *beep* *boop* --- <details><summary>launchdarkly-server-sdk-ai: 0.19.0</summary> ## [0.19.0](launchdarkly-server-sdk-ai-0.18.0...launchdarkly-server-sdk-ai-0.19.0) (2026-05-05) ### ⚠ BREAKING CHANGES * StructuredResponse replaced by RunnerResult with new "parsed" property * AgentResult replaced by RunnerResult and Managed Result * Removed ModelRunner and AgentRunner protocols * Removed invoke_method, invoke_structured_model from AIProvider base class. * ModelResponse was replaced by RunnerResult * Add ManagedResult, RunnerResult, and Runner protocol; rename invoke() to run() ([#148](#148)) * Swap track_metrics_of parameter order to match spec ([#144](#144)) ### Features * Add evaluations support to ManagedAgent.run() ([#153](#153)) ([442f46a](442f46a)) * Add judge evaluation support to agent graphs ([#142](#142)) ([3d5a6a9](3d5a6a9)) * Add ManagedGraphResult, GraphMetricSummary, and AgentGraphRunnerResult types ([#151](#151)) ([301e24c](301e24c)) * Add ManagedResult, RunnerResult, and Runner protocol; rename invoke() to run() ([#148](#148)) ([88d4ddc](88d4ddc)) * Add root-level tools map with customParameters to AI Config types ([#141](#141)) ([f17c535](f17c535)) * bake sampling_rate into Judge at construction; simplify Evaluator to List[Judge] ([#159](#159)) ([86c79e6](86c79e6)) * Update LangChain runners to implement Runner protocol returning RunnerResult ([#150](#150)) ([62a8e25](62a8e25)) ### Bug Fixes * Add runtime DeprecationWarnings to deprecated methods ([#145](#145)) ([2189b81](2189b81)) * AgentResult replaced by RunnerResult and Managed Result ([fbb0b4b](fbb0b4b)) * build judge input as string; strip legacy judge config messages ([#165](#165)) ([e6942a6](e6942a6)) * Fall back to model.parameters.tools when root tools absent ([#146](#146)) ([2c30d75](2c30d75)) * Graph tracking refactor — ManagedAgentGraph drives tracking for new runner shape ([#154](#154)) ([20a5020](20a5020)) * ModelResponse was replaced by RunnerResult ([fbb0b4b](fbb0b4b)) * parse model.parameters.tools as list ([#160](#160)) ([fb53e99](fb53e99)) * reference correct PyPI package names in provider load error messages ([#164](#164)) ([48761c9](48761c9)) * Removed invoke_method, invoke_structured_model from AIProvider base class. ([fbb0b4b](fbb0b4b)) * Removed ModelRunner and AgentRunner protocols ([fbb0b4b](fbb0b4b)) * Replace done_callback with coroutine chain for judge tracking ([#147](#147)) ([1e1f36b](1e1f36b)) * StructuredResponse replaced by RunnerResult with new "parsed" property ([fbb0b4b](fbb0b4b)) * Swap track_metrics_of parameter order to match spec ([#144](#144)) ([53db736](53db736)) </details> <details><summary>launchdarkly-server-sdk-ai-langchain: 0.6.0</summary> ## [0.6.0](launchdarkly-server-sdk-ai-langchain-0.5.0...launchdarkly-server-sdk-ai-langchain-0.6.0) (2026-05-05) ### Features * Add judge evaluation support to agent graphs ([#142](#142)) ([3d5a6a9](3d5a6a9)) * Migrate LangGraph runner to AgentGraphRunnerResult; clean up legacy shape detection ([#156](#156)) ([efa8e00](efa8e00)) * Support conversation history directly in AI Provider model runners ([#166](#166)) ([4bb3e78](4bb3e78)) * Update LangChain runners to implement Runner protocol returning RunnerResult ([#150](#150)) ([62a8e25](62a8e25)) ### Bug Fixes * build judge input as string; strip legacy judge config messages ([#165](#165)) ([e6942a6](e6942a6)) </details> <details><summary>launchdarkly-server-sdk-ai-openai: 0.5.0</summary> ## [0.5.0](launchdarkly-server-sdk-ai-openai-0.4.0...launchdarkly-server-sdk-ai-openai-0.5.0) (2026-05-05) ### Features * Add judge evaluation support to agent graphs ([#142](#142)) ([3d5a6a9](3d5a6a9)) * Support conversation history directly in AI Provider model runners ([#166](#166)) ([4bb3e78](4bb3e78)) * Update OpenAI graph runner to return AgentGraphRunnerResult with GraphMetrics ([#155](#155)) ([388b7af](388b7af)) * Update OpenAI runners to implement Runner protocol returning RunnerResult ([#149](#149)) ([382e662](382e662)) ### Bug Fixes * build judge input as string; strip legacy judge config messages ([#165](#165)) ([e6942a6](e6942a6)) </details> --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). <!-- CURSOR_SUMMARY --> --- > [!NOTE] > **Medium Risk** > Primarily a release/version bump, but it publishes **breaking API changes** (move to unified `Runner.run()`/`RunnerResult` and removal of `invoke_*` methods), which can break downstream integrations. > > **Overview** > Cuts a new release across the core SDK and provider packages: `launchdarkly-server-sdk-ai` to `0.19.0`, LangChain provider to `0.6.0`, and OpenAI provider to `0.5.0`, updating the release manifest and package metadata accordingly. > > Changelogs document the shipped breaking API surface changes (notably removing `invoke_model()`/`invoke_structured_model()` in favor of `run(...)` and standardizing returns on `RunnerResult`) plus accompanying feature/fix entries; the core package version constants/docs (`__version__`, `PROVENANCE.md`) are updated to match. > > <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit a20d7a5. Bugbot is set up for automated code reviews on this repo. Configure [here](https://www.cursor.com/dashboard/bugbot).</sup> <!-- /CURSOR_SUMMARY --> --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: jsonbailey <jbailey@launchdarkly.com> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.

Summary
OpenAIAgentGraphRunner_NodeMetricsAccumulator— a lightweight per-node metrics collector replacingLDAIConfigTrackerinside the runnerAgentGraphRunnerResultwith populatedGraphMetrics(path,duration_ms,usage,node_metrics)ManagedAgentGraph._flush_graph_tracking()from the result metricsManagedAgentGraph._flush_graph_tracking()extended to drive per-node tracking fromresult.metrics.node_metricsusing graph node tracker factoriesManagedAgentGraph.run()pipeline (tracking events now come from the managed layer)track_handoff_success()calls removed (per spec:pathfield is sufficient; handoffs are not inGraphMetrics)Depends on
Test plan
uv run pytest packages/ai-providers/server-ai-openai/tests/)test_openai_agent_graph_runner.py: runner returns new shape, no tracker createdtest_tracking_openai_agents.py: graph-level and per-node events emitted through managed layer🤖 Generated with Claude Code
Note
Medium Risk
Changes the runner result shape and shifts LaunchDarkly event emission from the OpenAI runner to
ManagedAgentGraph, which could affect downstream integrations and metric correctness if node/path/usage attribution is wrong.Overview
OpenAIAgentGraphRunnernow returnsAgentGraphRunnerResultwithGraphMetrics(path,duration_ms,usage, and per-nodenode_metrics) and removes all direct LaunchDarkly tracker calls, including handoff tracking.Per-node durations/usage/tool calls are accumulated inside the runner (
_node_metrics) during handoffs and final segment flush, and the integration path is updated soManagedAgentGraphemits graph + node tracking events by consumingresult.metrics.node_metricsand creating node trackers on demand.Tests are updated to assert the new result fields, ensure trackers are no longer created/called by the runner, and to run tracking assertions through
ManagedAgentGraph.run()(including failure cases and multi-node runs).Reviewed by Cursor Bugbot for commit fa936cf. Bugbot is set up for automated code reviews on this repo. Configure here.