Litellm ishaan april15

[ishaan-berri]

Relevant issues

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/test_litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Screenshots / Proof of Fix

Type

🆕 New Feature
🐛 Bug Fix
🧹 Refactoring
📖 Documentation
🚄 Infrastructure
✅ Test

Changes

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Apr 16, 2026 1:17am

[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 3 committers have signed the CLA.

✅ kothamah
❌ ishaan-berri
❌ harish876
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[greptile-apps]

Greptile Summary

This PR delivers two independent bug fixes: (1) a null-safety patch for Bedrock Guardrails, replacing .get("key", []) with .get("key") or [] at every list-field access to handle the Bedrock API returning explicit JSON null (which .get("key", []) does not guard against), and (2) a performance optimization for get_all_latest_health_checks that replaces a full-table Python-side dedup with a DB-level DISTINCT ON (model_id, model_name) ORDER BY checked_at DESC, backed by a new composite index added via a single-statement CREATE INDEX CONCURRENTLY migration (correctly kept single-statement so Prisma runs it outside a transaction).

Confidence Score: 5/5

Safe to merge — all changes are well-tested and no P0/P1 issues were found.

The bedrock null-safety fixes are guarded by the existing top-level try/except and covered by thorough new tests. The health-check query optimization is semantically equivalent to the old Python-side dedup, the single-statement CONCURRENTLY migration is correctly designed to run outside Prisma's transaction wrapping, and all remaining test modifications accurately reflect the new DB-side deduplication contract rather than weakening coverage.

No files require special attention.

Important Files Changed

Filename	Overview
litellm/proxy/utils.py	Replaces full-table-scan + Python-side deduplication with Prisma distinct=["model_id","model_name"] ORDER BY checked_at DESC, eliminating O(n) memory usage; logic is semantically equivalent.
litellm/proxy/guardrails/guardrail_hooks/bedrock_guardrails.py	Replaces .get("key", []) with .get("key") or [] throughout to handle explicit null values returned by the Bedrock API; function-level try/except ensures graceful degradation.
litellm-proxy-extras/litellm_proxy_extras/migrations/20260415120000_health_check_latest_per_model_index/migration.sql	Adds composite index on (model_id, model_name, checked_at DESC) using CREATE INDEX CONCURRENTLY; correctly kept as a single statement so Prisma runs it outside a transaction.
tests/test_litellm/proxy/guardrails/guardrail_hooks/test_bedrock_guardrails.py	Adds comprehensive null-safety test classes covering all policy sub-list fields, including BLOCKED-detection through mixed null/non-null fields; tests are well-structured and use mocks correctly.
tests/test_litellm/proxy/test_health_check_functions.py	Existing tests updated to reflect DB-side deduplication (mock now returns already-distinct rows); new test covers same model_name with and without model_id returning two distinct DB rows.
scripts/health_check/benchmark_get_all_latest_health_checks.py	New benchmark script for measuring memory and latency of get_all_latest_health_checks; inserts a runtime sys.path manipulation to import from the repo root which is acceptable for a standalone script.

Sequence Diagram

sequenceDiagram
    participant Caller
    participant get_all_latest_health_checks
    participant PostgreSQL

    Note over get_all_latest_health_checks,PostgreSQL: NEW: DB-side deduplication
    Caller->>get_all_latest_health_checks: await get_all_latest_health_checks()
    get_all_latest_health_checks->>PostgreSQL: SELECT DISTINCT ON (model_id, model_name) *
FROM LiteLLM_HealthCheckTable
ORDER BY model_id, model_name, checked_at DESC
    PostgreSQL-->>get_all_latest_health_checks: N distinct rows (one per model)
    get_all_latest_health_checks-->>Caller: list of latest health checks

    Note over get_all_latest_health_checks,PostgreSQL: OLD: full table scan + Python dedup
    Caller->>get_all_latest_health_checks: await get_all_latest_health_checks()
    get_all_latest_health_checks->>PostgreSQL: SELECT * FROM LiteLLM_HealthCheckTable ORDER BY checked_at DESC
    PostgreSQL-->>get_all_latest_health_checks: ALL historical rows
    get_all_latest_health_checks->>get_all_latest_health_checks: Python dedup loop (O(n) memory)
    get_all_latest_health_checks-->>Caller: list of latest health checks

Loading

_{Reviews (2): Last reviewed commit: "fix(ci): sync uv.lock with pyproject.tom..." | Re-trigger Greptile}