Spaces:
Paused
Paused
File size: 5,401 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 | # -*- coding: utf-8 -*-
"""Example of OpenAI Chat model multimodal (vision) calls using DataBlock."""
import asyncio
import base64
import os
from pathlib import Path
from _utils import stream_and_collect
from agentscope.message import (
Msg,
TextBlock,
DataBlock,
URLSource,
Base64Source,
)
from agentscope.model import OpenAIChatModel
from agentscope.credential import OpenAICredential
# 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"
)
# A publicly accessible test audio
TEST_AUDIO_URL = (
"https://help-static-aliyun-doc.aliyuncs.com/file-manage"
"-files/zh-CN/20250211/tixcef/cherry.wav"
)
async def example_image_url() -> None:
"""Call gpt-4.1 with an image URL and ask what is in the image."""
model = OpenAIChatModel(
credential=OpenAICredential(
api_key=os.environ["OPENAI_API_KEY"],
),
model="gpt-4.1",
stream=True,
context_size=1_047_576,
)
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() -> OpenAIChatModel:
return OpenAIChatModel(
credential=OpenAICredential(api_key=os.environ["OPENAI_API_KEY"]),
model="gpt-4.1",
stream=True,
context_size=1_047_576,
)
async def example_image_local_path() -> None:
"""Call gpt-4.1 with a local image using a ``file://`` URL.
The formatter reads the file from disk and converts it to a base64 data
URI.
"""
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 gpt-4.1 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))
async def example_audio() -> None:
"""Call gpt-audio-mini with an audio URL.
Audio understanding requires an audio-capable model such as
``gpt-audio-mini``. The formatter converts the audio source
to the ``input_audio`` format expected by the Chat Completions API.
"""
model = OpenAIChatModel(
credential=OpenAICredential(
api_key=os.environ["OPENAI_API_KEY"],
),
model="gpt-audio-mini",
stream=True,
)
audio_block = DataBlock(
source=URLSource(
url=TEST_AUDIO_URL,
media_type="audio/wav",
),
)
msgs = [
Msg(
name="user",
content=[
TextBlock(text="What is being said in this audio clip?"),
audio_block,
],
role="user",
),
]
print("=== Multimodal Call (Audio Input and Output) ===")
response = await stream_and_collect(
await model(
msgs,
modalities=["text", "audio"],
audio={"voice": "alloy", "format": "pcm16"},
),
)
# Save audio if present
for block in response.content:
if isinstance(block, DataBlock) and block.source.media_type.startswith(
"audio/",
):
audio_bytes = base64.b64decode(block.source.data)
print(f" Audio received: {len(audio_bytes)} bytes")
if __name__ == "__main__":
asyncio.run(example_image_url())
asyncio.run(example_image_local_path())
asyncio.run(example_image_base64())
asyncio.run(example_audio())
|