scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-03 02:22:24 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 82

fix: skip non-OpenAI file blocks in message updates

Browse source 
Updated the `update_messages_with_model_file_ids` and `get_file_ids_from_messages` functions to skip content blocks with type='file' that do not conform to the OpenAI Chat Completions shape. This prevents unhandled KeyErrors and allows for smoother processing of messages with varying file block structures.

Added regression tests to ensure that non-compliant file blocks are handled gracefully without raising errors.
This commit is contained in:
S0ngRu1 2026-05-13 15:35:48 +08:00
parent c2231b8af2
commit 264df382d4
2 changed files with 163 additions and 30 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

20
litellm/litellm_core_utils/prompt_templates/common_utils.py
View file

@ -470,11 +470,12 @@ def update_messages_with_model_file_ids(
file_object = cast(ChatCompletionFileObject, c)
file_object_file_field = file_object.get("file")
if not isinstance(file_object_file_field, dict):
raise litellm.BadRequestError(
message="Content block has type='file' but is missing the required 'file' field",
model=None,
llm_provider=None,
)
# Content block has `type: "file"` but not the
# OpenAI Chat Completions shape (e.g. a LangChain
# v1 standardized file block, or a provider-native
# shape that also uses `type: "file"`). Nothing to
# remap here, so skip instead of crashing.
continue
file_id = file_object_file_field.get("file_id")
format = file_object_file_field.get(
"format", get_format_from_file_id(file_id)
@ -1109,11 +1110,10 @@ def get_file_ids_from_messages(messages: List[AllMessageValues]) -> List[str]:
file_object = cast(ChatCompletionFileObject, c)
file_object_file_field = file_object.get("file")
if not isinstance(file_object_file_field, dict):
raise litellm.BadRequestError(
message="Content block has type='file' but is missing the required 'file' field",
model=None,
llm_provider=None,
)
# Content block has `type: "file"` but not the
# OpenAI Chat Completions shape. No file_id to
# extract, so skip instead of raising KeyError.
continue
file_id = file_object_file_field.get("file_id")
if file_id:
file_ids.append(file_id)

173
tests/test_litellm/llms/test_file_content_block.py
View file

@ -1,14 +1,17 @@
"""
Tests for handling malformed 'file' content blocks (missing 'file' sub-field).
Tests for handling malformed or invalid 'file' content blocks (missing or null
`file` sub-field, HTTP file_id URLs for Google AI Studio).
Regression tests for:
- litellm/llms/vertex_ai/gemini/transformation.py
- litellm/llms/gemini/chat/transformation.py
- litellm/litellm_core_utils/prompt_templates/common_utils.py
(migrate_file_to_image_url raises on missing `file`; file-id helpers skip non-OpenAI shapes)
- litellm/litellm_core_utils/prompt_templates/factory.py (Bedrock + Anthropic)
- litellm/llms/openai/chat/gpt_transformation.py
"""
import asyncio
import copy
from typing import List, cast
@ -62,6 +65,11 @@ MALFORMED_FILE_OBJECT: ChatCompletionFileObject = cast(
ChatCompletionFileObject, {"type": "file"}
)
EXPLICIT_NULL_FILE_OBJECT: ChatCompletionFileObject = cast(
ChatCompletionFileObject,
{"type": "file", "file": None},
)
def _malformed() -> List[AllMessageValues]:
return copy.deepcopy(cast(List[AllMessageValues], _MALFORMED_MESSAGES_RAW))
@ -71,6 +79,23 @@ def _well_formed() -> List[AllMessageValues]:
return copy.deepcopy(cast(List[AllMessageValues], _WELL_FORMED_MESSAGES_RAW))
def _explicit_null_file_in_content() -> List[AllMessageValues]:
return copy.deepcopy(
cast(
List[AllMessageValues],
[
{
"role": "user",
"content": [
{"type": "text", "text": "hello"},
{"type": "file", "file": None},
],
}
],
)
)
# ---------------------------------------------------------------------------
# vertex_ai/gemini/transformation.py
# ---------------------------------------------------------------------------
@ -86,6 +111,15 @@ def test_gemini_convert_messages_malformed_file_raises_bad_request():
)
def test_gemini_convert_messages_explicit_null_file_field_raises_bad_request():
"""Explicit JSON null for `file` must be rejected like a missing `file` key."""
with pytest.raises(litellm.BadRequestError, match="missing the required 'file' field"):
_gemini_convert_messages_with_history(
messages=_explicit_null_file_in_content(),
model="gemini-2.0-flash",
)
# ---------------------------------------------------------------------------
# gemini/chat/transformation.py - GoogleAIStudioGeminiConfig
# ---------------------------------------------------------------------------
@ -99,20 +133,78 @@ def test_google_ai_studio_transform_messages_malformed_file_raises_bad_request()
config._transform_messages(messages=_malformed(), model="gemini-2.0-flash")
def test_google_ai_studio_transform_messages_explicit_null_file_field_raises_bad_request():
"""Explicit JSON null for `file` must be rejected like a missing `file` key."""
config = GoogleAIStudioGeminiConfig()
with pytest.raises(litellm.BadRequestError, match="missing the required 'file' field"):
config._transform_messages(
messages=_explicit_null_file_in_content(), model="gemini-2.0-flash"
)
def test_google_ai_studio_transform_messages_http_file_id_converts_to_base64(monkeypatch):
"""Google AI Studio rejects raw HTTP(S) file URLs; _transform_messages should
fetch and replace them with base64 `file_data` before conversion."""
# Data URL shape so downstream Gemini media parsing accepts the inlined bytes
# (mirrors real `convert_url_to_base64` output from `_process_image_response`).
fake_file_data = "data:application/pdf;base64,aGVsbG8="
def _fake_convert_url_to_base64(url: str) -> str:
assert url == "https://example.com/doc.pdf"
return fake_file_data
monkeypatch.setattr(
"litellm.llms.gemini.chat.transformation.convert_url_to_base64",
_fake_convert_url_to_base64,
)
messages = cast(
List[AllMessageValues],
[
{
"role": "user",
"content": [
{"type": "text", "text": "hello"},
{
"type": "file",
"file": {
"file_id": "https://example.com/doc.pdf",
"format": "pdf",
},
},
],
}
],
)
config = GoogleAIStudioGeminiConfig()
config._transform_messages(messages=messages, model="gemini-2.0-flash")
content = messages[0].get("content")
assert isinstance(content, list)
file_block = next(c for c in content if isinstance(c, dict) and c.get("type") == "file")
file_field = file_block.get("file")
assert isinstance(file_field, dict)
assert file_field.get("file_data") == fake_file_data
assert "file_id" not in file_field
# ---------------------------------------------------------------------------
# common_utils.py - update_messages_with_model_file_ids
# ---------------------------------------------------------------------------
def test_update_messages_with_model_file_ids_malformed_raises_bad_request():
"""update_messages_with_model_file_ids should raise BadRequestError for content
blocks that have type='file' but no 'file' sub-field."""
with pytest.raises(litellm.BadRequestError, match="missing the required 'file' field"):
update_messages_with_model_file_ids(
messages=_malformed(),
model_id="some-model",
model_file_id_mapping={},
)
def test_update_messages_with_model_file_ids_malformed_skips_non_openai_file_block():
"""Non-OpenAI file blocks (e.g. missing nested `file` dict) are skipped so callers
relying on LangChain v1 / provider-native shapes are not rejected here."""
messages = _malformed()
result = update_messages_with_model_file_ids(
messages=messages,
model_id="some-model",
model_file_id_mapping={},
)
assert result == messages
content = result[0].get("content")
assert isinstance(content, list)
file_block = next(c for c in content if isinstance(c, dict) and c.get("type") == "file")
assert "file" not in file_block
def test_update_messages_with_model_file_ids_well_formed_updates():
@ -134,10 +226,9 @@ def test_update_messages_with_model_file_ids_well_formed_updates():
# ---------------------------------------------------------------------------
def test_get_file_ids_from_messages_malformed_raises_bad_request():
"""get_file_ids_from_messages should raise BadRequestError for malformed file blocks."""
with pytest.raises(litellm.BadRequestError, match="missing the required 'file' field"):
get_file_ids_from_messages(messages=_malformed())
def test_get_file_ids_from_messages_malformed_skips_non_openai_file_block():
"""Blocks with type='file' but no OpenAI `file` sub-dict yield no extracted ids."""
assert get_file_ids_from_messages(messages=_malformed()) == []
def test_get_file_ids_from_messages_well_formed_returns_ids():
@ -170,14 +261,36 @@ def test_bedrock_process_file_message_malformed_raises_bad_request():
BedrockConverseMessagesProcessor._process_file_message(MALFORMED_FILE_OBJECT)
@pytest.mark.asyncio
async def test_bedrock_async_process_file_message_malformed_raises_bad_request():
def test_bedrock_process_file_message_explicit_null_file_field_raises_bad_request():
with pytest.raises(litellm.BadRequestError, match="missing the required 'file' field"):
BedrockConverseMessagesProcessor._process_file_message(EXPLICIT_NULL_FILE_OBJECT)
def test_bedrock_async_process_file_message_malformed_raises_bad_request():
"""_async_process_file_message should raise BadRequestError (not KeyError)
when the file object is missing the 'file' sub-field."""
with pytest.raises(litellm.BadRequestError, match="missing the required 'file' field"):
await BedrockConverseMessagesProcessor._async_process_file_message(
MALFORMED_FILE_OBJECT
)
async def _run() -> None:
with pytest.raises(
litellm.BadRequestError, match="missing the required 'file' field"
):
await BedrockConverseMessagesProcessor._async_process_file_message(
MALFORMED_FILE_OBJECT
)
asyncio.run(_run())
def test_bedrock_async_process_file_message_explicit_null_file_field_raises_bad_request():
async def _run() -> None:
with pytest.raises(
litellm.BadRequestError, match="missing the required 'file' field"
):
await BedrockConverseMessagesProcessor._async_process_file_message(
EXPLICIT_NULL_FILE_OBJECT
)
asyncio.run(_run())
# ---------------------------------------------------------------------------
@ -196,6 +309,16 @@ def test_openai_apply_common_transform_malformed_file_raises_bad_request():
config._apply_common_transform_content_item(malformed_block)
def test_openai_apply_common_transform_explicit_null_file_field_raises_bad_request():
config = OpenAIGPTConfig()
explicit_null_block: OpenAIMessageContentListBlock = cast(
OpenAIMessageContentListBlock,
{"type": "file", "file": None},
)
with pytest.raises(litellm.BadRequestError, match="missing the required 'file' field"):
config._apply_common_transform_content_item(explicit_null_block)
def test_openai_apply_common_transform_well_formed_file_does_not_raise():
"""_apply_common_transform_content_item should not raise for well-formed file blocks."""
config = OpenAIGPTConfig()
@ -221,6 +344,11 @@ def test_anthropic_process_openai_file_message_malformed_raises_bad_request():
anthropic_process_openai_file_message(MALFORMED_FILE_OBJECT)
def test_anthropic_process_openai_file_message_explicit_null_file_field_raises_bad_request():
with pytest.raises(litellm.BadRequestError, match="missing the required 'file' field"):
anthropic_process_openai_file_message(EXPLICIT_NULL_FILE_OBJECT)
def test_anthropic_process_openai_file_message_well_formed_file_id_does_not_raise():
"""anthropic_process_openai_file_message should not raise for a well-formed file_id block."""
well_formed: ChatCompletionFileObject = cast(
@ -243,6 +371,11 @@ def test_migrate_file_to_image_url_malformed_raises_bad_request():
migrate_file_to_image_url(MALFORMED_FILE_OBJECT)
def test_migrate_file_to_image_url_explicit_null_file_field_raises_bad_request():
with pytest.raises(litellm.BadRequestError, match="missing the required 'file' field"):
migrate_file_to_image_url(EXPLICIT_NULL_FILE_OBJECT)
def test_migrate_file_to_image_url_well_formed_returns_image_url():
"""migrate_file_to_image_url should return an image_url block for a well-formed file."""
well_formed: ChatCompletionFileObject = cast(