refactor(utils): remove the dead get_api_key provider-key resolver

[devin-ai-integration]

TLDR

Problem this solves:

• litellm.utils.get_api_key has no callers anywhere in the repo
• It duplicates provider key resolution that get_llm_provider already does
• Being unobservable, it hid a misspelled env var for years (fix(ai21): resolve the documented AI21_API_KEY instead of a misspelled name #35985)
• An unused import in main.py also published it as litellm.get_api_key

How it solves it:

• Delete the function, the unused main.py import, and its ai21 test
• Ratchet the ruff, type-discipline and basedpyright budgets it was consuming

User Flow

Before: nothing a proxy or SDK user does is affected, the function is dead code. The only visible effect of this PR is on a Python caller who reached the function by import

1. A Python script runs import litellm; litellm.get_api_key("openai", None)
2. It returns the resolved key, so the script keeps working even though the function only knows 12 of the 100+ providers and returns None for the rest

After: the same import fails loudly instead of returning a partial answer

1. The same script runs import litellm; litellm.get_api_key("openai", None)
2. It raises AttributeError: module 'litellm' has no attribute 'get_api_key'
3. Every proxy route and SDK call (completion, get_llm_provider, ...) behaves exactly as before, see the live run below

Relevant issues

Follow-up to #35985, which fixed the misspelling but deliberately left the function in place so the removal could get its own review

Linear ticket

Resolves LIT-5245

Pre-Submission checklist

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

• I have added meaningful tests (n/a: this PR only deletes code, and a test asserting a name is absent would be a structure test; the deleted ai21 test was the only one that referenced the function)
• The handful of test files covering my change pass locally, e.g. uv run pytest tests/test_litellm/<your_test_file>.py -v (tests/test_litellm/test_utils.py: 317 passed)
• My PR passes all required CI/CD checks (e.g., lint, schema.d.ts sync check, etc.)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 (5/5 at 11a7a4d) before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Delays in PR merge?

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

Screenshots / Proof of Fix

Shared setup: PYTHONPATH=$PWD so the interpreter and the proxy load this checkout, litellm/proxy/dev_config.yaml unchanged, Postgres running, ANTHROPIC_API_KEY set in the environment

Before (92d4533)

Import surface

1. OPENAI_API_KEY=sk-proof-only uv run --no-sync python -c 'import inspect, litellm; print(inspect.getsourcefile(litellm.get_api_key), inspect.getsourcelines(litellm.get_api_key)[1]); print(litellm.get_api_key("openai", None)); print(litellm.get_api_key("bedrock", None))'
2. Output:

/home/ubuntu/repos/litellm/litellm/utils.py 5109
sk-proof-only
None

The function resolves at the top level of the package through main.py's unused import, answers for openai, and silently returns None for bedrock and every other provider it does not list

After (11a7a4d)

Import surface

1. OPENAI_API_KEY=sk-proof-only uv run --no-sync python -c 'import litellm, litellm.utils; print(hasattr(litellm, "get_api_key"), hasattr(litellm.utils, "get_api_key"), litellm.completion.__module__)'
2. Output:

False False litellm.main

Live proxy still serves a real completion

1. uv run --no-sync litellm --config litellm/proxy/dev_config.yaml --port 4000 boots to GET /health/liveliness -> 200
2. curl -s http://localhost:4000/chat/completions -H "Authorization: Bearer sk-1234" -H "Content-Type: application/json" -d '{"model":"anthropic-haiku-4-5","messages":[{"role":"user","content":"Reply with exactly: get_api_key removed"}],"max_tokens":20}'
3. Output (real Anthropic call, 17 prompt tokens and 9 completion tokens billed):

{"id":"chatcmpl-82de9809-3892-4d79-9708-fe93f49f4379","created":1788319769,"model":"anthropic-haiku-4-5","object":"chat.completion","choices":[{"finish_reason":"stop","index":0,"message":{"content":"get_api_key removed","role":"assistant","provider_specific_fields":{"citations":null,"thinking_blocks":null}}}],"usage":{"completion_tokens":9,"prompt_tokens":17,"total_tokens":26,...}}

4. curl -s http://localhost:4000/models -H "Authorization: Bearer sk-1234" lists all 40 configured deployments, so provider resolution for the whole model list still works without the deleted helper

Type

🧹 Refactoring

Caveats (if any)

Medium

• litellm.get_api_key and litellm.utils.get_api_key stop resolving for external importers
  • Undocumented, not in __init__.py, published only via main.py's unused import and the from .main import * in __init__.py
  • Precedent for removing such a name outright rather than deprecating first: refactor(bedrock): remove the dead BedrockLLM invoke code path #35188 deleted BedrockLLM from the same star-exported import block
  • If the maintainers prefer deprecate-first, the alternative is a DeprecationWarning shim in utils.py and removal on a major

Low

• The wider problem the ticket surfaced, main.py having no __all__ so litellm.* gains or loses names whenever an import changes, is out of scope here and should get its own ticket
• The non-required osv-scan check is red on this PR and on the last three litellm_internal_staging runs alike (pypdf and tornado advisories in the untouched uv.lock), so it is inherited from the base, not introduced here

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

Link to Devin session: https://app.devin.ai/sessions/7d2f5ce02d6a4968925246a56c43960c
Open in Devin Desktop: https://app.devin.ai/desktop/session/7d2f5ce02d6a4968925246a56c43960c?variant=devin
Requested by: @yassin-berriai

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

@greptileai please review this dead-code removal, in particular the public surface caveat on litellm.get_api_key

[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.}

[devin-ai-integration]

PR #39260 (author devin-ai-integration[bot], repo BerriAI/litellm) carries no labels — the enterprise label is absent, so no risk label, product label, or Linear change was made. Exiting without mutations.

[codspeed]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing litellm_remove_dead_get_api_key (11a7a4d) with litellm_internal_staging (92d4533)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[devin-ai-integration]

@greptileai review

[greptile-apps]

Greptile Summary

This PR removes the unused get_api_key resolver and its accidental package-level export.

• Deletes the resolver and its AI21-specific test.
• Removes the unused import from litellm/main.py.
• Tightens static-analysis budgets to reflect the deleted code.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
litellm/utils.py	Removes the dead provider-key resolver as intended.
litellm/main.py	Removes the import that accidentally exposed the resolver at package level.
tests/test_litellm/test_utils.py	Removes the test dedicated solely to the deleted resolver.
basedpyright-code-budget.json	Ratchets type-checking budgets to account for the deleted code.
ruff-strict-budget.json	Ratchets lint budgets to account for the removed function and import.
type-discipline-budget.json	Lowers the affected type-discipline budget consistently with the deletion.

_{Reviews (2): Last reviewed commit: "refactor(utils): remove the dead get_api..." | Re-trigger Greptile}

[devin-ai-integration]

@greptileai re-review please, the compatibility thread above is resolved and the code is unchanged at 11a7a4d