Walkthrough

Deployment

Kubrick includes a CLI tool that automates the process of setting up and deploying the infrastructure, meaning developers can have an ingestion pipeline up and running in a matter of minutes.

git clone https://github.com/kubrick-ai/kubrick.git
cd kubrick/cli && npm run build
kubrick deploy

Full CLI deployment documentation is available here.

API

Deploying the infrastructure will expose the Kubrick API:

bash

curl -X POST https://API_URL/v1_0/search
-H Content-Type: multipart/form-data \
-F query_type=text \
-F query_text="A red sports car driving down the highway" \

js

const url = "https://API_URL/v1_0/search";
const formData = new FormData();
formData.append("query_type", "text");
formData.append("query_text", "A red sports car driving down the highway");

const response = await fetch(url, { method: "POST", body: formData });
const data = await response.json();

python

import requests

url = "https://API_URL/v1_0/search"
response = requests.post(
    url,
    data={
        "query_type": "text",
        "query_text": "A red sports car driving down the highway",
    },
)
data = response.json()

Example Response:

{
  "data": [
    {
      "id": 1234,
      "modality": "visual-text",
      "scope": "clip",
      "start_time": 6.0,
      "end_time": 12.0,
      "similarity": 0.7654,
      "video": {
        "id": 5678,
        "s3_bucket": "kubrick-video-library-<uuid>",
        "s3_key": "uploads/<uuid>/red-sports-car.mp4",
        "filename": "red-sports-car.mp4",
        "duration": 120.0,
        "created_at": "2025-08-05T00:00:00.000000",
        "updated_at": "2025-08-05T00:00:00.000000",
        "height": 1080,
        "width": 1920,
        "url": "<presigned_url>"
      }
    }
    ...
  ],
  "metadata": { "page": 0, "limit": 10, "total": 10 }
}

Full API documentation is available here.

Playground

The deployment includes a playground web interface, hosted as an S3 static site. The playground serves as an implementation example for the core features of the Kubrick API.

Ingestion

Users can upload videos into a private library, which will automatically create an embedding task.

Text Queries

Text queries return relevant results even if they describe qualities that aren’t concrete like emotion, tone, composition, visual style, and movement.

Media Queries

Kubrick also enables querying with image, audio, and video files. Users can upload a reference image to find visually similar scenes, search using an audio clip to locate moments with matching soundscapes or music, or provide a video segment to retrieve the most similar clips in the library.

Image

Video/Audio