[litellm-agent] Staging → litellm_internal_staging (5/6/2026)

[oss-pr-review-agent-shin]

Automated staging PR created by litellm-agent.

This branch collects PRs approved by the agent on 5/6/2026.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ pnookala-godaddy
❌ oss-pr-review-agent-shin[bot]
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[greptile-apps]

Greptile Summary

This PR applies a consistent sorted iteration order to all spend-update transaction loops across the proxy's DB write layer, so concurrent pods acquire row locks in the same order and avoid PostgreSQL deadlocks. A parametrized mock test verifies the sorted order for all 8 spend buckets (user, key, team, team_member, org, end_user, tag, agent).

• Six loops in db_spend_update_writer.py and one in utils.py now call sorted(transactions.items()) before iterating, with inline comments explaining the deadlock-prevention rationale.
• The composite key format for team_member_list_transactions ("team_id::<v>::user_id::<v>") means a lexicographic string sort is equivalent to a (team_id, user_id) tuple sort, which is correctly noted in the comment.
• The new test is mock-only and consistent with the repository's CI rule for the tests/test_litellm/ folder.

Confidence Score: 5/5

This PR is safe to merge — it changes only iteration order within existing transactions, not the data written, and all paths are covered by mock tests.

The change is a targeted, low-risk addition of sorted() to dictionary iteration in spend-flush loops. The sort is deterministic, doesn't alter the values written, and is backed by a comprehensive parametrized test suite. The composite-key sort assumption for team_member entries is correct given the fixed-length prefix format. No new DB paths, no API changes, and no real network calls in the tests.

No files require special attention.

Important Files Changed

Filename	Overview
litellm/proxy/db/db_spend_update_writer.py	Adds sorted() to 5 spend-update transaction loops (user, key, team, team_member, org) and the shared _update_entity_spend_in_db helper to enforce a consistent row-lock acquisition order and prevent cross-pod PostgreSQL deadlocks.
litellm/proxy/utils.py	Applies the same sorted() fix to the end_user spend loop in ProxyUpdateSpend.update_end_user_spend, completing deadlock prevention coverage across all spend buckets.
tests/test_litellm/proxy/db/test_db_spend_update_writer.py	Adds a parametrized mock-only test covering all 8 spend buckets to assert that _commit_spend_updates_to_db iterates each bucket in sorted order; no real network calls are made.

_{Reviews (1): Last reviewed commit: "Merge PR #27221 into agent staging branc..." | Re-trigger Greptile}

[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: Merged into staging branch litellm_agent_oss_staging_05_06_2026. Staging PR: #27250

---

Triage Summary
Sorts dictionary iteration order by ID before issuing batched spend-update writes in db_spend_update_writer.py and proxy/utils.py, ensuring consistent lock acquisition order across pods to prevent database deadlocks. Adds 143 lines of parametrized tests covering user, key, team, team_member, and org transaction buckets to verify sorted ordering is preserved.

Merge Confidence: 5/5 ✅ READY
Ready to ship.

All checks green. Greptile 5/5, no blocking pattern findings, CircleCI passed.