fix(otel): normalise list guardrail_mode to tuple in _emit_once to prevent TypeError

[devteamaegis]

Problem

When a guardrail is configured with a list-valued mode (e.g. mode: ["pre_call", "post_call"]) and the OpenTelemetry integration is active, every request that triggers the guardrail returns HTTP 500:

TypeError: unhashable type: 'list'

Bisected to #27757 (commit 1c4e4d4).

Closes #28486.

Root cause

_emit_once constructs a deduplification key:

dedupe_key = (self.__class__.__name__, id(self), *scope)

scope is unpacked via *, so if one element is a list (the guardrail_mode), the resulting tuple contains a list. Tuples that contain unhashable elements (like list) are themselves unhashable — so spans_logged.get(dedupe_key) raises TypeError.

Fix

Normalise each scope part before building the key: convert list values to tuple (which is hashable and preserves identity for dedupe purposes):

normalized_scope = tuple(
    tuple(part) if isinstance(part, list) else part
    for part in scope
)
dedupe_key = (self.__class__.__name__, id(self), *normalized_scope)

All non-list scope parts pass through unchanged (strings, ints, floats, None, etc. are all hashable).

Changes

• litellm/integrations/opentelemetry.py: normalise scope parts in _emit_once
• tests/test_litellm/integrations/test_opentelemetry.py: add test_guardrail_mode_as_list_does_not_crash

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[greptile-apps]

Greptile Summary

Fixes a TypeError: unhashable type: 'list' crash in _emit_once that occurred whenever a guardrail was configured with a list-valued mode (e.g. ["pre_call", "post_call"]) and the OpenTelemetry integration was active, causing every such request to return HTTP 500.

• litellm/integrations/opentelemetry.py: Before constructing the deduplication tuple key, each element of scope is normalised — lists are converted to tuples (which are hashable), while all other types pass through unchanged.
• tests/test_litellm/integrations/test_opentelemetry.py: Adds a mock-only regression test test_guardrail_mode_as_list_does_not_crash that exercises the exact failure path and confirms the fix resolves it.

Confidence Score: 4/5

Safe to merge — the change is a one-line normalisation in a well-contained helper with a direct regression test covering the failure path.

The fix is correct and minimal. The only notable gap is that the new regression test doesn't assert mock_span.end.assert_called_once(), which the other guardrail tests do include, leaving an unchecked span-lifecycle branch.

No files require special attention beyond the minor missing assertion in the new test.

Important Files Changed

Filename	Overview
litellm/integrations/opentelemetry.py	Normalises list-valued scope parts to tuples in _emit_once so the deduplication key is always hashable; fix is minimal and correctly targeted at the root cause.
tests/test_litellm/integrations/test_opentelemetry.py	Adds a focused regression test for the list-mode crash; uses mocks only (no network calls), but omits mock_span.end.assert_called_once() that other guardrail tests include.

_{Reviews (1): Last reviewed commit: "test(otel): add regression for list guar..." | Re-trigger Greptile}

[greptile-apps]

The new test verifies that start_span is called but doesn't assert that the span was properly ended. The sibling test test_create_guardrail_span_with_valid_info includes mock_span.end.assert_called_once(). Without that assertion here, a future regression where the span is created but never closed would go undetected by this test.

Suggested change

	# Must not raise TypeError: unhashable type: 'list'
	otel._create_guardrail_span(kwargs=kwargs, context=None)
	otel.tracer.start_span.assert_called_once()
	# Must not raise TypeError: unhashable type: 'list'
	otel._create_guardrail_span(kwargs=kwargs, context=None)
	otel.tracer.start_span.assert_called_once()
	mock_span.end.assert_called_once()

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[oss-pr-review-agent-shin]

🤖 litellm-agent: This PR is currently BLOCKED from merge.

Score: 3/5 ❌

Why blocked:

• 1 PR-related CI failure (lint) (pr_related_failures, -2 pts)

Details: Score docked for: 1 PR-related CI failure (lint).

Fix the issues above and push an update — the bot will re-review automatically.

Note: This bot is still in beta and might not always work as expected. Please share any feedback via Slack.

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.