Some answers are in a document’s pictures, not its text. PageIndex Cloud saves a document’s images when it indexes it, and the agent can fetch one and show it to a model that can see images.
Ask what the image on page 5 of a pump manual shows, in two steps:
- Ask the question.
- Look at the image the agent fetched.
Set up
pip install -U pageindexThe chat model has to read images, so pick a vision-capable one:
import os
from pageindex import PageIndexClient
os.environ["PAGEINDEX_API_KEY"] = "pageindex_demo_readonly_key"
os.environ["OPENAI_API_KEY"] = "Your OpenAI API key here"
client = PageIndexClient(index="cloud", chat="gpt-5.6-luna")
doc_id = "pi-cmu6jnfbb001k09ntgqmw5swr" # a pump manual on the demo accountAsk about the image
for chunk in client.chat("Describe the image in page 5", doc_id=doc_id, stream=True):
print(chunk, end="", flush=True)[tool_call] get_document_structure {"doc_name": "49eeaa61-....pdf", "folder_id": "cmu6jms8l000004l3cpmp7k6c"}
[tool_call] get_page_content {"doc_name": "49eeaa61-....pdf", "folder_id": "cmu6jms8l000004l3cpmp7k6c", "pages": "5"}
[tool_call] get_document_image {"image_path": "49eeaa61-....pdf/img-7.jpeg", "folder_id": "cmu6jms8l000004l3cpmp7k6c"}
Page 5 shows a schematic of the Hydra-Cell pump installation:
- The pump is positioned to the right, with a **pressure gauge** and **pressure regulator** on the discharge side.
- A **suction line** runs from the pump toward the inlet/supply tank.
- A **vacuum/pressure gauge** monitors the suction line.
- Fluid exits through the **discharge line** at the top left.
- A **bypass line** returns fluid from the pressure-regulator area to the inlet tank.
- Arrows indicate the flow direction through the suction, discharge, and bypass lines.The agent read page 5 and found that it has an image, img-7.jpeg. It fetched the image with get_document_image and described what it saw.
Show the image
get_document_image() returns a link to the same image:
from IPython.display import Image
Image(url=client.get_document_image(doc_id, "img-7.jpeg"))
Every labeled part of the drawing is in the description.
Result
The answer comes from the picture, not the text around it.
The notebook runs this with your key: open it in Colab .
Last updated on