Spaces:
Paused
Paused
File size: 3,676 Bytes
0b9dc2e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 | # -*- coding: utf-8 -*-
"""Example of Ollama model multimodal (vision) calls using DataBlock.
Note: Vision input requires a multimodal Ollama model such as llava, bakllava,
or moondream. Run `ollama pull llava` before executing this example.
"""
import asyncio
import base64
from pathlib import Path
from _utils import stream_and_collect
from agentscope.message import (
Msg,
TextBlock,
DataBlock,
URLSource,
Base64Source,
)
from agentscope.model import OllamaChatModel
# A publicly accessible test image (a simple cat photo)
TEST_IMAGE_URL = (
"https://help-static-aliyun-doc.aliyuncs.com/file-manage"
"-files/zh-CN/20241022/emyrja/dog_and_girl.jpeg"
)
async def example_image_url() -> None:
"""Call llava:7b with an image URL and ask what is in the image.
Requires `ollama pull llava` to be run first.
"""
model = OllamaChatModel(
model="llava:7b",
stream=True,
context_size=4_096,
)
image_block = DataBlock(
source=URLSource(
url=TEST_IMAGE_URL,
media_type="image/jpeg",
),
)
msgs = [
Msg(
name="user",
content=[
TextBlock(
text="What animal is in this image? Describe it briefly.",
),
image_block,
],
role="user",
),
]
print("=== Multimodal Call (Image URL) ===")
await stream_and_collect(await model(msgs))
def _build_model() -> OllamaChatModel:
return OllamaChatModel(
model="llava:7b",
stream=True,
context_size=4_096,
)
async def example_image_local_path() -> None:
"""Call llava:7b with a local image using a ``file://`` URL.
The formatter reads the file from disk and converts it to base64.
"""
model = _build_model()
abs_path = str(Path(__file__).parent / "test.jpeg")
msgs = [
Msg(
name="user",
content=[
TextBlock(
text="What is happening in this image? Describe it "
"briefly.",
),
DataBlock(
source=URLSource(
url=f"file://{abs_path}",
media_type="image/jpeg",
),
),
],
role="user",
),
]
print("=== Local Path Call (file://) ===")
await stream_and_collect(await model(msgs))
async def example_image_base64() -> None:
"""Call llava:7b with a local image using explicit base64 encoding.
Use ``Base64Source`` when you already have the binary data in memory or
want full control over the encoding step.
"""
model = _build_model()
with open(Path(__file__).parent / "test.jpeg", "rb") as f:
data = base64.b64encode(f.read()).decode("utf-8")
msgs = [
Msg(
name="user",
content=[
TextBlock(
text="What is happening in this image? Describe it "
"briefly.",
),
DataBlock(
source=Base64Source(data=data, media_type="image/jpeg"),
),
],
role="user",
),
]
print("=== Explicit Base64 Call ===")
await stream_and_collect(await model(msgs))
if __name__ == "__main__":
asyncio.run(example_image_url())
asyncio.run(example_image_local_path())
asyncio.run(example_image_base64())
|