31 lines
1.1 KiB
Python
31 lines
1.1 KiB
Python
|
"""Tests for is_vision_model (issue #124).
|
||
|
|
|
||
|
|
Local vision models served through Ollama/llama.cpp show up under many
|
||
|
|
names. If one isn't recognized as vision-capable, the image attachment is
|
||
|
|
stripped from the request before it reaches the model, so it silently never
|
||
|
|
sees the picture.
|
||
|
|
"""
|
||
|
|
from src.chat_helpers import is_vision_model
|
||
|
|
|
||
|
|
|
||
|
|
def test_recognizes_local_and_hosted_vision_models():
|
||
|
|
for name in [
|
||
|
|
# the ones #124 missed
|
||
|
|
"moondream", "moondream:latest",
|
||
|
|
"llama3.2-vision:11b", "granite3.2-vision",
|
||
|
|
"qwen2.5-vl:7b", "qwen2.5vl", "internvl2.5", "cogvlm",
|
||
|
|
# already worked, keep them working
|
||
|
|
"llava", "llava:7b", "bakllava", "minicpm-v",
|
||
|
|
"gpt-4o", "claude-sonnet-4", "gemini-2.0-flash", "pixtral-12b",
|
||
|
|
]:
|
||
|
|
assert is_vision_model(name), f"{name!r} should be detected as vision-capable"
|
||
|
|
|
||
|
|
|
||
|
|
def test_text_only_models_not_flagged():
|
||
|
|
for name in ["qwen2.5:3b", "mistral", "llama3.1:8b", "deepseek-r1", "phi3", "vicuna", ""]:
|
||
|
|
assert not is_vision_model(name), f"{name!r} should not be flagged as vision"
|
||
|
|
|
||
|
|
|
||
|
|
def test_none_is_safe():
|
||
|
|
assert is_vision_model(None) is False
|