Home

Duplicate Video Search

Tattle Khoj supports searching via images: https://services.tattle.co.in/khoj/ The search surfaces stories from Fact checking sites if they contain images similar to the one being searched for.

'Duplicate-video' is a sub project to implement Tattle video search on Khoj. This project will surface fact checking sites if they contain a video which is an exact match to the one searched for.

Tech Stack Used:

• Python and Flask for Backend
• React for Frontend
• SQL or Mongo for persistence
• Redis for caching needs

Dependencies on Other Tattle Projects:

For now this is a headless backend without any UI. When we integrate it as the search engine for khoj, the UI will talk to the archive-server (which handles authentication and routing to correct services), and archive-server will request this server to get results.

API Documentation

The server will be given a URL to a video file and not the file itself. The server might need to download the file if it needs to operate on the file.

1. Upload Endpoint (/api/upload) - This endpoint will be used to add video to our server.

Sample Code to upload a video to the server (via url)

import requests

url = "SERVER_URL/upload"

payload = { 
	"url": IMAGE_URL,
        "type" : "video"
	"doc_id": DOCUMENT_ID
}

response = requests.request("POST", url, data = payload)

print(response.text.encode('utf8'))

2. Search Endpoint (/api/search) - Sample Code to search for a duplicate video (via url)

This will return matching doc ids of the duplicates

import requests

url = "SERVER_URL/search"

payload = { 
	"url": IMAGE_URL,
        "type" : "video"
}

response = requests.request("POST", url, data = payload)

print(response.text.encode('utf8'))

Timelines: