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: count video_url content blocks

Browse source
This commit is contained in:
pragnyanramtha 2026-05-17 01:24:57 +00:00
parent 1b0ae3af83
commit b883652907
2 changed files with 78 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
litellm/litellm_core_utils/token_counter.py
View file

@ -723,6 +723,13 @@ def _count_content_list(
num_tokens += _count_image_tokens(
image_url, use_default_image_token_count
)
elif c["type"] == "video_url":
video_url = c.get("video_url", "")
if isinstance(video_url, dict):
video_url = video_url.get("url")
if video_url is None:
video_url = ""
num_tokens += DEFAULT_IMAGE_TOKEN_COUNT + count_function(str(video_url))
elif c["type"] in ("tool_use", "tool_result"):
num_tokens += _count_anthropic_content(
c,
@ -744,7 +751,7 @@ def _count_content_list(
)
raise ValueError(
f"Invalid content item type: {content_type}. "
f"Expected str or dict with 'type' field (text, image_url, tool_use, tool_result, thinking)."
f"Expected str or dict with 'type' field (text, image_url, video_url, tool_use, tool_result, thinking)."
)
return num_tokens
except Exception as e:

70
tests/test_litellm/litellm_core_utils/test_token_counter.py
View file

@ -16,6 +16,7 @@ from unittest.mock import AsyncMock, MagicMock, patch
import litellm
from litellm import create_pretrained_tokenizer, decode, encode, get_modified_max_tokens
from litellm import token_counter as token_counter_old
from litellm.constants import DEFAULT_IMAGE_TOKEN_COUNT
from litellm.litellm_core_utils.token_counter import token_counter as token_counter_new
from tests.large_text import text
from tests.test_litellm.litellm_core_utils.messages_with_counts import (
@ -908,6 +909,75 @@ def test_token_counter_with_image_url():
), f"Expected detail validation error, got: {e}"
def test_token_counter_with_video_url():
messages_dict = [
{
"role": "user",
"content": [
{"type": "text", "text": "Please describe this video."},
{
"type": "video_url",
"video_url": {
"url": "data:video/mp4;base64,AAAA",
},
},
],
}
]
tokens_dict = token_counter(model="gpt-4o", messages=messages_dict)
assert tokens_dict > 0, f"Expected positive token count, got {tokens_dict}"
messages_str = [
{
"role": "user",
"content": [
{
"type": "video_url",
"video_url": "https://example.com/video.mp4",
}
],
}
]
tokens_str = token_counter(model="gpt-4o", messages=messages_str)
assert (
tokens_str > 0
), f"Expected positive token count for string video_url, got {tokens_str}"
assert (
tokens_str > DEFAULT_IMAGE_TOKEN_COUNT
), f"Expected default video token budget, got {tokens_str}"
messages_empty_url = [
{
"role": "user",
"content": [{"type": "video_url", "video_url": ""}],
}
]
messages_none_url = [
{
"role": "user",
"content": [{"type": "video_url", "video_url": None}],
}
]
messages_none_nested_url = [
{
"role": "user",
"content": [{"type": "video_url", "video_url": {"url": None}}],
}
]
tokens_empty_url = token_counter(model="gpt-4o", messages=messages_empty_url)
assert tokens_empty_url > DEFAULT_IMAGE_TOKEN_COUNT
assert (
token_counter(model="gpt-4o", messages=messages_none_url) == tokens_empty_url
)
assert (
token_counter(model="gpt-4o", messages=messages_none_nested_url)
== tokens_empty_url
)
def test_token_counter_with_thinking_content():
"""
Test that _count_content_list() correctly handles Claude's extended thinking content blocks.