ColPali and Vespa Visual Retrieval

ColPali and Vespa Visual Retrieval

Demonstrates visual retrieval built on ColPali models and Vespa, as a Gradio app you can open online or run on your own computer. For search developers and researchers.

Made by vespa-engine on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space is a demo of visual retrieval that combines ColPali models with Vespa. The page gives only a short title, so its details come from the files in the repository. It names two ColPali model repositories and holds a backend, a frontend and static assets.

Who it suits: It suits developers and researchers who want to see how a ColPali-based retriever is wired to Vespa. The code is there to read and run, and the repository is large at 836 files.

Title: Who might use it
Industry | Who uses it | How they use it | Example
Search engineering | Developers | Study how a visual retriever connects to Vespa | Reading main.py and the backend folder
Document search | Researchers | Try ColPali-based retrieval in a working demo | Opening the Space and running a search
AI education | Teachers and students | Use the demo as a worked example | Walking through the project files in a class

People who want a documented product with settings and support can skip it, since the page offers only a title.

Where to use it:

• A web browser: open the Space page on Hugging Face
• Hugging Face Spaces: it runs there on a basic CPU machine
• Windows, macOS or Linux: run your own copy with Python 3.11

Title: Jobs it does
Use case | Who | What they get back
Trying visual retrieval | Researchers | A working demo to run searches in
Learning the setup | Developers | A project with its backend, frontend and static files to read
Local testing | Engineers | A copy running on their own computer

How to set it up: Using the online Space needs nothing installed. To run a copy you need these first:

• Python 3.11 on Windows, macOS or Linux
• The packages in requirements.txt, with Gradio 4.44.0

1. Install the Hugging Face hub package.
2. Download the Space into a folder named colpali-vespa-visual-retrieval.
3. Move into that folder.
4. Install Gradio 4.44.0, then the other packages.
5. Start the app.

pip install -U huggingface_hub
hf download vespa-engine/colpali-vespa-visual-retrieval --repo-type space --local-dir colpali-vespa-visual-retrieval
cd colpali-vespa-visual-retrieval
pip install gradio==4.44.0
pip install -r requirements.txt
python main.py

How to use it: Online, open the Space page and use the search it shows. On your own computer, start the app with this command and open it the same way.

python main.py

The source documents no other commands or settings.

Example: You start the app and run a search for a document page you are looking for. The title says retrieval is visual, so expect matches chosen by how pages look. The page shows no sample queries or results.

For newcomers and experts: If you are new to this, think of it as a search demo for documents that treats pages as pictures. Open the Space page and try one search to see what comes back.

If you work in the field, the page names vidore/colpaligemma-3b-pt-448-base and vidore/colpali-v1.2 as its models. The app is set up in Hugging Face as Gradio 4.44.0 with main.py as its main file. The repository holds backend and frontend folders plus styling files, so read them to see how retrieval and Vespa fit together.

Open the Space or run main.py
Enter a search in the app
ColPali models and Vespa handle retrieval
Results appear in the page

Good to know: The README is a single title line, so the code is the main documentation.

• License: apache-2
• Cost to run: none named
• Data it reaches: the source does not say, so check the code before sending private documents

On Hugging Face, 13 November 2026

• Likes: 65
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/vespa-engine/colpali-vespa-visual-retrieval
• Files: https://huggingface.co/spaces/vespa-engine/colpali-vespa-visual-retrieval/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.