[avara1986]

Description

Jira: APPSEC-70768

The AI Guard evaluate response can now carry an optional extra_tags object (data.attributes.extra_tags), so new evaluation details reach AI Guard spans without a tracer change per tag.

• Scalar values (str, int, float, bool) are set as tags on the ai_guard span with the key unchanged. Booleans are written as "true"/"false", matching the other AI Guard tags.
• Structured values (dict, list) are added to the ai_guard meta_struct under their key.
• Values the tracer computes are never overridden: ai_guard.action, reason, target, tool_name, blocked, redacted, event, the anomaly-detection copies, and the meta_struct keys messages, attack_categories, sds, tag_probs.
• Keys that Span.set_tag treats as span controls are also reserved: manual.keep, manual.drop, service.name, service.version, _dd.measured, version, env.
• Backend tags may override customer tags passed through the tags= parameter.
• Tags are applied before the block decision, so they land on both the allow and blocking paths.
• Fail-safe: a malformed payload (not a dict, non-string or empty keys, unsupported value types) is skipped with a debug log and never changes the evaluation result or telemetry.

No public API change.

Testing

• New tests/aiguard/api/test_extra_tags.py (15 tests): scalars, structured values, reserved tracer and control keys, anomaly-tag reservation, customer-tag override, blocking path, malformed and null payloads, absent field, and an exception raised while applying tags.
• ai_guard_api suite passes on py3.9, 3.11 and 3.13; ai_guard_openai suite passes on py3.13.
• Lint (format, style, typing, spelling, security), plus import cycle and layering checks: no new findings.

Risks

Low. The new path only runs when the response contains extra_tags, and it is wrapped so that it cannot fail the evaluation. The response field name follows the expected backend contract; if that changes, the parsing is a one-line update.

Additional Notes

Backend values are not size-capped; they come from decoded JSON returned by the AI Guard service.

🤖 Generated with Claude Code

[cit-pr-commenter-54b7da]

Codeowners resolved as

Resolved from the full PR diff against main using the target branch CODEOWNERS file.
CODEOWNERS team requests not listed below are not required by the current file set.

ddtrace/aiguard/_api_client.py                                          @DataDog/asm-python
tests/aiguard/api/test_extra_tags.py                                    @DataDog/asm-python
tests/aiguard/utils.py                                                  @DataDog/asm-python

[cit-pr-commenter-54b7da]

Dependency direction analysis

⚠️ Existing dependency direction violations

There are 163 dependency direction violations that already exist on the base branch and have not been changed by this PR.

Show existing violations (showing 5 of 163 highest severity)

ddtrace.internal.tracemethods -×-> ddtrace.trace  (internal-core -> product:tracing, score=121)
ddtrace.llmobs._integrations.openai -×-> ddtrace.trace  (product:llmobs -> product:tracing, score=119)
ddtrace.llmobs._integrations.bedrock_agents -×-> ddtrace.trace  (product:llmobs -> product:tracing, score=119)
ddtrace.llmobs._utils -×-> ddtrace.trace  (product:llmobs -> product:tracing, score=119)
ddtrace.llmobs._integrations.langgraph -×-> ddtrace.trace  (product:llmobs -> product:tracing, score=119)

To see all violations, download the layers-base.json and layers-pr.json artifacts from this CI job and run:

uv run --script scripts/import-analysis/layers.py compare layers-base.json layers-pr.json

[avara1986]

@codex review

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-21T13:00:44.589988Z	4f05433	New commits

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Another round soon, please!

Reviewed commit: 16b5696a5b

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".

[datadog-prod-us1-3]

More details

Reserved span controls and tracer-owned AI Guard fields remain protected, and extra tags are applied before the block decision. Validation used source inspection and isolated checks; the test suite was not run.

Was this helpful? React 👍 or 👎

_{Open Bits AI session}

_{🤖 Bits Code Review · Commit 247e644}

[florentinl]

LGTM !

Just a tiny optional typing nit.