Content

Content is the input used to generate a response from an LLM. Supported content types are:

• TEXT
• IMAGE
• DOCUMENT
• VIDEO

To use IMAGE, DOCUMENT, or VIDEO input, make sure to use a multi-modal model that supports the type.

---

Text

Text content can be passed directly as a string without the Content class:

response = llmfy.invoke("Hello")

Or as a list of Content objects:

from llmfy import Content

content = [
    Content(value="Hello"),
]

response = llmfy.invoke(content)

---

Image

Image input is supported by OpenAI, AWS Bedrock, and Google AI.

OpenAI

Accepts a base64 data URI or an image URL.

import base64
from llmfy import Content, ContentType

input_image = "path/to/image.jpg"
with open(input_image, "rb") as f:
    image = f"data:image/jpeg;base64,{base64.b64encode(f.read()).decode('utf-8')}"

# Or use an image URL directly:
# image = "https://example.com/image.jpg"

content = [
    Content(value="Describe this image."),
    Content(type=ContentType.IMAGE, value=image),
]

response = llmfy.invoke(content)

AWS Bedrock

Accepts raw image bytes or an S3 URI. Supported formats: gif, jpeg, png, webp.

Using bytes:

from llmfy import Content, ContentType

input_image = "path/to/image.jpg"
with open(input_image, "rb") as f:
    image_bytes = f.read()

content = [
    Content(value="Describe this image."),
    Content(
        type=ContentType.IMAGE,
        format="jpeg",
        value=image_bytes,
    ),
]

response = llmfy.invoke(content)

Using AWS S3:

content = [
    Content(value="Describe this image."),
    Content(
        type=ContentType.IMAGE,
        use_s3=True,
        bucket_owner="111122223333",
        value="s3://amzn-s3-demo-bucket/myImage",
    ),
]

response = llmfy.invoke(content)

Google AI

Accepts a base64 data URI or an HTTP URL.

import base64
from llmfy import Content, ContentType

input_image = "path/to/image.jpg"
with open(input_image, "rb") as f:
    image = f"data:image/jpeg;base64,{base64.b64encode(f.read()).decode('utf-8')}"

# Or use an HTTP URL directly:
# image = "https://example.com/image.jpg"

content = [
    Content(value="Describe this image."),
    Content(type=ContentType.IMAGE, value=image),
]

response = llmfy.invoke(content)

---

Document

Document (PDF only) input is supported by OpenAI, AWS Bedrock, and Google AI.

OpenAI

Accepts a base64 data URI. The filename must include the .pdf extension.

import base64
from llmfy import Content, ContentType

input_doc = "path/to/document.pdf"
with open(input_doc, "rb") as f:
    doc = f"data:application/pdf;base64,{base64.b64encode(f.read()).decode('utf-8')}"

content = [
    Content(
        type=ContentType.DOCUMENT,
        filename="document.pdf",
        value=doc,
    ),
    Content(value="Who are the characters in this document?"),
]

response = llmfy.invoke(content)

AWS Bedrock

Accepts raw PDF bytes or an S3 URI.

Using bytes:

from llmfy import Content, ContentType

input_doc = "path/to/document.pdf"
with open(input_doc, "rb") as f:
    doc = f.read()

content = [
    Content(
        type=ContentType.DOCUMENT,
        filename="document",
        value=doc,
    ),
    Content(
        type=ContentType.TEXT,
        value="Who are the characters in this document?",
    ),
]

response = llmfy.invoke(content)

Using AWS S3:

content = [
    Content(
        type=ContentType.DOCUMENT,
        use_s3=True,
        bucket_owner="111122223333",
        value="s3://amzn-s3-demo-bucket/myPdf",
    ),
    Content(
        type=ContentType.TEXT,
        value="Who are the characters in this document?",
    ),
]

response = llmfy.invoke(content)

Google AI

Accepts a base64 data URI. The filename must include the .pdf extension.

import base64
from llmfy import Content, ContentType

input_doc = "path/to/document.pdf"
with open(input_doc, "rb") as f:
    doc = f"data:application/pdf;base64,{base64.b64encode(f.read()).decode('utf-8')}"

content = [
    Content(
        type=ContentType.DOCUMENT,
        filename="document.pdf",
        value=doc,
    ),
    Content(value="Who are the characters in this document?"),
]

response = llmfy.invoke(content)

---

Video

Video input is supported by AWS Bedrock and Google AI. OpenAI does not support video input yet.

OpenAI

Not supported yet.

AWS Bedrock

Accepts raw video bytes or an S3 URI. Supported formats: wmv, mpg, mpeg, three_gp, flv, mp4, mov, mkv, webm.

Using bytes:

from llmfy import Content, ContentType

input_video = "path/to/video.mp4"
with open(input_video, "rb") as f:
    video_bytes = f.read()

content = [
    Content(type=ContentType.TEXT, value="What happens in this video?"),
    Content(
        type=ContentType.VIDEO,
        format="mp4",
        value=video_bytes,
    ),
]

response = llmfy.invoke(content)

Using AWS S3:

content = [
    Content(type=ContentType.TEXT, value="What happens in this video?"),
    Content(
        type=ContentType.VIDEO,
        use_s3=True,
        bucket_owner="111122223333",
        value="s3://amzn-s3-demo-bucket/myVideo",
    ),
]

response = llmfy.invoke(content)

Google AI

Accepts an HTTP URL, a YouTube URL, or raw video bytes (recommended for files under 20 MB).

Using a URL:

from llmfy import Content, ContentType

content = [
    Content(type=ContentType.TEXT, value="What happens in this video?"),
    Content(
        type=ContentType.VIDEO,
        value="https://example.com/video.mp4",
    ),
]

response = llmfy.invoke(content)

Using bytes (< 20 MB):

from llmfy import Content, ContentType

input_video = "path/to/video.mp4"
with open(input_video, "rb") as f:
    video_bytes = f.read()

content = [
    Content(type=ContentType.TEXT, value="What happens in this video?"),
    Content(
        type=ContentType.VIDEO,
        value=video_bytes,
        format="mp4",  # optional, defaults to mp4
    ),
]

response = llmfy.invoke(content)