feat(google-genai): capture Interactions request configuration by JacksonWeber · Pull Request #752 · open-telemetry/opentelemetry-python-genai

JacksonWeber · 2026-09-21T00:38:27Z

Description

Captures explicitly supplied generation configuration and output-format attributes for Google GenAI Interactions calls, including sync, async, streaming, and failed requests.

These settings were previously omitted from telemetry. The implementation supports legacy keyword arguments and newer SDK request.body models, preserves zero values, and leaves SDK arguments and lazy iterators untouched.

Closes #563

Field availability follows the SDK version. Mixed output formats are omitted when no single semantic-convention output type represents them.

Assisted-by: GPT-6 Astra

Type of change

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to not work as expected)
This change requires a documentation update

How has this been tested?

tox -e precommit,typecheck
Google oldest, v2, and latest suites, including Python 3.14 latest.
Google conformance scenarios with Weaver 0.26.1 and VCR replay.
Offline SDK-model and HTTP-transport coverage for parameters, streaming normalization, and error paths. No live Gemini calls or API key.

Checklist

See CONTRIBUTING.md
for the style guide, changelog guidance, and more.

Followed the style guidelines of this project
Changelog updated if the change requires an entry
Unit tests added
Documentation updated

Record generation configuration and output-format attributes for sync, async, and streaming Interactions calls, including newer request.body inputs. Preserve SDK stream normalization, zero values, and lazy request content. Closes open-telemetry#563 Assisted-by: GPT-6 Astra Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

opentelemetry-pr-dashboard · 2026-09-21T00:39:55Z

Pull request dashboard status

Merged · refreshed 2026-09-22 05:40 UTC

Status above doesn't look right?

Anything look wrong? Report it with what you expected; it helps us improve the dashboard.

Copilot

Copilot review overview

🟡 Changes recommended

The request-body extraction currently uses __dict__, which can include defaulted/unset model fields and may record configuration that was not explicitly supplied.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1

Open (1)

Exclude defaulted Pydantic fields from recorded request attributes · New

What changed in this PR

This PR extends the Google GenAI Interactions instrumentation to capture explicitly supplied request configuration (generation parameters and output format) as gen_ai.request.* and gen_ai.output.type attributes across sync/async/streaming and error paths, and adds focused tests and documentation for the new telemetry.

Changes:

Extract request configuration from both legacy kwargs and newer request.body shapes, and apply it onto the util-genai invocation before completion/stream finalization.
Derive gen_ai.output.type from response_format / response_mime_type when representable by a single semconv output type.
Add a comprehensive Interactions request-parameter test suite plus README documentation for the new behavior.

File	Description
instrumentation/opentelemetry-instrumentation-google-genai/src/opentelemetry/instrumentation/google_genai/interactions.py	Adds request-shape normalization + attribute extraction for generation config and output type; adjusts stream detection to work with request-body models.
instrumentation/opentelemetry-instrumentation-google-genai/tests/interactions/test_request_parameters.py	New tests covering sync/async/streaming/error paths, request-body vs kwargs shapes, and output-type mapping.
instrumentation/opentelemetry-instrumentation-google-genai/README.rst	Documents which Interactions parameters are captured and when output type is set/omitted.
instrumentation/opentelemetry-instrumentation-google-genai/.changelog/0.added	Adds a changelog fragment describing the new captured attributes.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

Replace the placeholder fragment number with the assigned upstream PR number required by the changelog workflow. Assisted-by: GPT-6 Astra Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Use Pydantic field metadata to capture only explicitly supplied request and generation-config fields, including explicit extras. Avoid model serialization so lazy content and SDK argument objects remain untouched. Assisted-by: GPT-6 Astra Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Copilot review overview

🟢 Approval recommended

The implementation is behavior-preserving, semconv-aligned, and comprehensively tested.

Review effort: Balanced
Findings: None

Resolved since last review (1)

Exclude defaulted Pydantic fields from recorded request attributes

lmolkova · 2026-09-21T04:29:09Z

 If both variables are set, the includes list is applied first, then the
 excludes list filters the result further.

+Interactions request parameters


Can we de-AI this section a bit? I think it can be either removed (we don't list the details for each operation) or collapsed to 1-2 sentences.

Makes sense, removed this section to clean this up

lmolkova · 2026-09-21T04:41:31Z

+    if format_type in (
+        "object",
+        "array",
+        "string",
+        "number",
+        "integer",
+        "boolean",
+        "null",
+    ):
+        return output_types.JSON.value


Suggested change

if format_type in (

"object",

"array",

"string",

"number",

"integer",

"boolean",

"null",

):

return output_types.JSON.value

if format_type in (

"object",

"array",

"string",

"number",

"integer",

"boolean",

"null",

"json",

"json_object",

"json_schema",

):

return output_types.JSON.value

Might be worth including common JSON format names too.

Added json, json_object, and json_schema here and updated the test request params

lmolkova · 2026-09-21T04:41:48Z

+    for name in ("temperature", "top_p"):
+        value = _get_field(config, name)
+        if (
+            isinstance(value, (int, float))
+            and not isinstance(value, bool)
+            and -float_info.max <= value <= float_info.max
+        ):
+            setattr(invocation, name, float(value))


Suggested change

for name in ("temperature", "top_p"):

value = _get_field(config, name)

if (

isinstance(value, (int, float))

and not isinstance(value, bool)

and -float_info.max <= value <= float_info.max

):

setattr(invocation, name, float(value))

for name in ("temperature", "top_p"):

value = _get_field(config, name)

if (

isinstance(value, (int, float))

and not isinstance(value, bool)

and math.isfinite(value)

and value >= 0

):

setattr(invocation, name, float(value))

math.isfinite could be simpler than checking against float_info.

Updated to use math.isfinite

lmolkova · 2026-09-21T04:42:02Z

+    if isinstance(stop_sequences, (list, tuple)) and all(
+        isinstance(item, str) for item in stop_sequences
+    ):
+        invocation.stop_sequences = list(stop_sequences)


Suggested change

if isinstance(stop_sequences, (list, tuple)) and all(

isinstance(item, str) for item in stop_sequences

):

invocation.stop_sequences = list(stop_sequences)

if (

isinstance(stop_sequences, (list, tuple))

and stop_sequences

and all(isinstance(item, str) for item in stop_sequences)

):

invocation.stop_sequences = list(stop_sequences)

Consider checking that stop_sequences is non-empty so [] is not recorded as an attribute.

Added a check on stop_sequences.

eternalcuriouslearner · 2026-09-21T12:49:29Z

Sorry overlooked one small thing, should we add assertions to verify these attributes to vcr tests and conformance test suite?

Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…/opentelemetry/instrumentation/google_genai/interactions.py Co-authored-by: Liudmila Molkova <neskazu@gmail.com>

Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

lmolkova · 2026-09-21T16:58:06Z

+    for name in ("temperature", "top_p"):
+        value = _get_field(config, name)
+        if isinstance(value, int) and value > float_info.max:
+            continue


btw why do we even need to protect from it, if top_p is invalid float, the SDK would throw, why do we need to validate boundaries?

We convert the value before calling the SDK, so a huge integer could make our instrumentation throw first. This guard just skips recording it and lets the SDK handle the original value. Although it's definitely an unlikely edge case.

lmolkova · 2026-09-21T17:00:06Z

+    request: object,
+) -> None:
+    config = _explicit_request_fields(_get_field(request, "generation_config"))
+    for name in ("temperature", "top_p"):


why do we have a loop here? please copy over method _coerse_float from some other place in this repo and call it for individual properties.

And the let's use typed invocatio.top_p | temperature setters.

Moved the existing conversion checks into _coerce_float and replaced the loop with explicit invocation.temperature and invocation.top_p assignments

lmolkova · 2026-09-21T17:00:54Z

    telemetry_handler: TelemetryHandler,
    instance: InteractionsResource | AsyncInteractionsResource,
-    kwargs: dict[str, Any],
+    request: object,


is there a specific type we can use instead of object?

Changed this to Mapping[str, object] and updated the request helpers to match.

JacksonWeber · 2026-09-21T17:24:58Z

Sorry overlooked one small thing, should we add assertions to verify these attributes to vcr tests and conformance test suite?

Added assertions to the sync/async VCR tests and the conformance scenario.

…tests Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

JacksonWeber requested a review from a team as a code owner September 21, 2026 00:38

Copilot AI lite review requested due to automatic review settings September 21, 2026 00:38

github-project-automation Bot added this to GenAI Semantic Conventions and Instrumentation libraries Sep 21, 2026

Copilot started reviewing on behalf of JacksonWeber September 21, 2026 00:39 View session

opentelemetry-pr-dashboard Bot mentioned this pull request Sep 21, 2026

Pull Request Dashboard #171

Open

Copilot AI reviewed Sep 21, 2026

View reviewed changes

Comment thread ...-instrumentation-google-genai/src/opentelemetry/instrumentation/google_genai/interactions.py Outdated

JacksonWeber and others added 2 commits September 21, 2026 00:42

fix: number Google GenAI changelog fragment for PR 752

497c2a3

Replace the placeholder fragment number with the assigned upstream PR number required by the changelog workflow. Assisted-by: GPT-6 Astra Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

eternalcuriouslearner requested a balanced review from Copilot September 21, 2026 03:45

Copilot started reviewing on behalf of eternalcuriouslearner September 21, 2026 03:45 View session

Copilot AI reviewed Sep 21, 2026

View reviewed changes

eternalcuriouslearner approved these changes Sep 21, 2026

View reviewed changes

lmolkova reviewed Sep 21, 2026

View reviewed changes

JacksonWeber and others added 6 commits September 21, 2026 09:07

docs(google-genai): remove interactions request parameter details

0000b4e

Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Update instrumentation/opentelemetry-instrumentation-google-genai/src…

8f86751

…/opentelemetry/instrumentation/google_genai/interactions.py Co-authored-by: Liudmila Molkova <neskazu@gmail.com>

fix(google-genai): recognize common JSON response formats

4d4f873

Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

fix(google-genai): validate finite sampling parameters

4a68334

Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

fix(google-genai): omit empty stop sequences

a071030

Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Merge branch 'main' into feat/google-interactions-config

d099ba9

lmolkova reviewed Sep 21, 2026

View reviewed changes

JacksonWeber and others added 4 commits September 21, 2026 10:25

test(google-genai): assert request attributes in VCR and conformance …

30fc38d

…tests Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

refactor(google-genai): use typed sampling parameter assignments

bcb09c3

Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

refactor(google-genai): type normalized request mappings

9d9ea0b

Assisted-by: GPT-6 Astra Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Merge branch 'main' into feat/google-interactions-config

5aac52b

JacksonWeber requested a review from lmolkova September 21, 2026 18:23

lmolkova approved these changes Sep 22, 2026

View reviewed changes

lmolkova added this pull request to the merge queue Sep 22, 2026

Merged via the queue into open-telemetry:main with commit 6f69095 Sep 22, 2026
78 checks passed

github-project-automation Bot moved this to Done in GenAI Semantic Conventions and Instrumentation libraries Sep 22, 2026

Conversation

JacksonWeber commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Type of change

How has this been tested?

Checklist

Uh oh!

opentelemetry-pr-dashboard Bot commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pull request dashboard status

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟢 Approval recommended

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

eternalcuriouslearner commented Sep 21, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

JacksonWeber commented Sep 21, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

JacksonWeber commented Sep 21, 2026 •

edited

Loading

opentelemetry-pr-dashboard Bot commented Sep 21, 2026 •

edited

Loading