Add Agent Framework reference scenario

[eavanvalkenburg]

Changes

• Add a native Microsoft Agent Framework reference scenario using OpenAIChatClient.
• Exercise native AF telemetry for agent invocation, model inference, and function tool execution.
• Extend the local OpenAI Responses mock with Responses-style function-call output and usage-detail fields so AF can emit current GenAI usage attributes.
• Refresh reference reports and the reference README.

Notes

This scenario intentionally reflects what Agent Framework emits natively today. It does not add manual reference spans for surfaces AF does not currently emit, such as GenAI workflow spans or inference operation detail events.

[Copilot]

Pull request overview

This PR adds a new Python reference scenario for the Microsoft Agent Framework, exercising its native GenAI telemetry (agent invocation, inference, and tool execution) against the repo’s local OpenAI-compatible mock server. It also extends the mock server’s /responses behavior to support Responses-style tool calling and usage detail fields so the Agent Framework can emit current GenAI usage attributes, and refreshes the generated reference reports/README.

Changes:

• Added reference/scenarios/agent-framework/ (scenario script, dependency metadata, lockfile, and generated data.json) to capture native Agent Framework telemetry.
• Extended the mock OpenAI Responses endpoint to return function-call outputs and richer usage details (cached_tokens, reasoning_tokens).
• Refreshed generated coverage reports and the reference README to include agent-framework as a supporting library.

Reviewed changes

Copilot reviewed 9 out of 10 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
reference/src/semconv_genai/mock_server/openai.py	Adds Responses usage details and a tool-call response path for /responses.
reference/src/semconv_genai/mock_server/_common.py	Broadens tool-argument schema extraction to support additional tool shapes.
reference/scenarios/agent-framework/scenario.py	New scenario exercising Agent Framework native telemetry for agent runs, inference, and tools.
reference/scenarios/agent-framework/pyproject.toml	Defines dependencies for the new Agent Framework scenario.
reference/scenarios/agent-framework/uv.lock	Locks scenario dependencies for reproducible runs.
reference/scenarios/agent-framework/data.json	Captured span→attribute coverage emitted by the scenario run.
reference/reports/invoke-agent-internal-span.md	Regenerated report including agent-framework coverage.
reference/reports/inference-span.md	Regenerated report including agent-framework coverage (incl. usage detail attrs).
reference/reports/execute-tool-span.md	Regenerated report including agent-framework tool execution coverage.
reference/README.md	Updates the scenario/report index to list agent-framework as supported.

[otelbot]

Copilot has reviewed this PR. Copilot's suggestions aren't always correct or applicable, so please evaluate each comment on its merits and then handle it in one of these ways:

• click GitHub's "Apply suggestion" button (auto-resolves the thread);
• reply that it was applied (ideally linking to the commit); or
• reply that it was not applied, with the reason; or
• reply with a question for reviewers.

Automation flags a PR for human review once every Copilot comment has a reply or is marked as resolved, so keeping these threads up to date helps reviewers know when the PR is ready.

Status across open PRs is visible on the pull request dashboard.

[lmolkova]

Thanks! It seems ADK does not produce the workflow span, but produces inference ones. Inference span would likely be duplicated by the underlying openai instrumentation, we should try to avoid emitting them.

[eavanvalkenburg]

@lmolkova thanks, I fixed the pipeline and noticed some things missing, should be good now.

[lmolkova]

thanks @eavanvalkenburg !

JFYI - I'm proposing to remove all inference manual instrumentation from reference scenarios here #351 with some principles outlined in AGENTS.md - would love to know if you have any thoughts or concerns

Non-editorial contributions need two approvals, given it's documenting what Agent Framework de-facto does, I'm going to consider this PR to be editorial and merge it. Thanks!

[eavanvalkenburg]

@lmolkova i do have some concern, as far as I'm aware, we instrument everything in Agent framework and then we do use openai or Anthropic or others to call models, but I've never seen them emitting things, likely those libraries do not emit by default (I might have seen logs, but def not spans and metrics). So this change would require all libraries like ours to figure out for each provider how to set that up, or make the dev ex for our customers worse because they have to do two instrumentation setups, and if they want to switch provider it goes from a single line change up multiple updates to still get otel. So while I can follow the thinking in theory, make each thing responsible for its own features, the practice is not as simple. Related to #351