Technology Sharing

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

---

1. How to directly transfer multimodal data to the model

Here, we demonstrate how to pass multimodal input directly to the model. For other model providers that support multimodal input, langchain provides internal logic in the class to convert to the expected format.
The most common way to pass in an image is to pass it in as a byte string. This should work for most model ensembles.

import base64
import httpx

image_url = "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg"
image_data = base64.b64encode(httpx.get(image_url).content).decode("utf-8")

message = HumanMessage(
    content=[
        {"type": "text", "text": "describe the weather in this image"},
        {
            "type": "image_url",
            "image_url": {"url": f"data:image/jpeg;base64,{image_data}"},
        },
    ],
)
response = model.invoke([message])
print(response.content)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

We can provide image URLs directly in a content block of type "image_URL". Note however that only some model providers support this feature.

message = HumanMessage(
    content=[
        {"type": "text", "text": "describe the weather in this image"},
        {"type": "image_url", "image_url": {"url": image_url}},
    ],
)
response = model.invoke([message])
print(response.content)
1
2
3
4
5
6
7
8

We can also upload multiple pictures.

message = HumanMessage(
    content=[
        {"type": "text", "text": "are these two images the same?"},
        {"type": "image_url", "image_url": {"url": image_url}},
        {"type": "image_url", "image_url": {"url": image_url}},
    ],
)
response = model.invoke([message])
print(response.content)
1
2
3
4
5
6
7
8
9

---

2. How to use mutimodal prompts

Here, we'll describe how to use prompt templates to format multimodal imputs for models.

import base64
import httpx

image_url = "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg"
image_data = base64.b64encode(httpx.get(image_url).content).decode("utf-8")

prompt = ChatPromptTemplate.from_messages(
    [
        ("system", "Describe the image provided"),
        (
            "user",
            [
                {
                    "type": "image_url",
                    "image_url": {"url": "data:image/jpeg;base64,{image_data}"},
                }
            ],
        ),
    ]
)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

We can also pass multiple images to the model.

prompt = ChatPromptTemplate.from_messages(
    [
        ("system", "compare the two pictures provided"),
        (
            "user",
            [
                {
                    "type": "image_url",
                    "image_url": {"url": "data:image/jpeg;base64,{image_data1}"},
                },
                {
                    "type": "image_url",
                    "image_url": {"url": "data:image/jpeg;base64,{image_data2}"},
                },
            ],
        ),
    ]
)

chain = prompt | model

response = chain.invoke({"image_data1": image_data, "image_data2": image_data})
print(response.content)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23