fix(utils): normalize list content strings to OpenAI text parts

[lkapadiya-DO]

Relevant issues

Fixes an observed issue where sending an array of strings in the content block (e.g. ["what is the capital of France?"]) causes an internal AttributeError: 'str' object has no attribute 'get'. This previously bubbled up as a 500 Internal Server Error, confusingly attaching the extracted llm_provider to the crash response.

Also list content can include Pydantic BaseModel parts (not only strings/dicts), and invalid items (null, integers) should fail fast instead of passing through to downstream .get() calls.

Linear ticket

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

Local manual verification
From the repo root, run:

cd litellm
uv sync --dev

uv run python <<'PY'
from litellm.utils import validate_and_fix_openai_messages

out = validate_and_fix_openai_messages(
    [{"role": "user", "content": ["what is the capital of France?"]}]
)
print("Case A:", out[0]["content"])

out = validate_and_fix_openai_messages(
    [{"role": "user", "content": [
        "some text",
        {"type": "image_url", "image_url": {"url": "https://example.com/x.png"}},
    ]}]
)
print("Case B:", out[0]["content"])

try:
    validate_and_fix_openai_messages([{"role": "user", "content": [None]}])
except TypeError as e:
    print("Case D OK:", e)
PY

Expected output:

Case A: [{'type': 'text', 'text': 'what is the capital of France?'}]
Case B: [{'type': 'text', 'text': 'some text'}, {'type': 'image_url', 'image_url': {'url': 'https://example.com/x.png'}}]
Case D OK: Invalid message type: <class 'NoneType'>. Expected dict or Pydantic model.

Before:
Sending an array of strings (e.g. "content": ["what is the capital of France?"]) resulted in an internal crash that returned a 500 error containing internal proxy metadata:

{
  "error": {
    "message": "'str' object has no attribute 'get'",
    "type": "server_error",
    "param": null,
    "code": null,
    "llm_provider": "openai"
  }
}

After:
The array of strings is normalized into the OpenAI multimodal dict format before downstream processing:
"content": [{"type": "text", "text": "what is the capital of France?"}]

(Tests passing successfully locally — pytest + manual script above)

Type

🐛 Bug Fix

Changes

1. Updated validate_and_fix_openai_messages in litellm/utils.py to normalize list content for all messages (after tool_calls handling, before convert_to_dict):

• str → {"type": "text", "text": item}
• dict → unchanged (e.g. existing image_url parts)
• other (e.g. Pydantic BaseModel) → convert_to_dict(item) via model_dump()
• invalid items (null, int, etc.) → TypeError at validation time (fail fast)

2. Added pytest coverage in tests/test_litellm/test_utils.py:

• test_normalize_array_of_strings_in_content
• test_normalize_pydantic_content_part_in_content
• test_normalize_invalid_content_item_raises_type_error (parametrized)