GuidesSet up

Create a knowledge store

A knowledge store holds your videos and the understanding the platform derives from them. Create one before you can query your content through the Responses API. For details on the underlying structure, see Knowledge stores.

To create a knowledge store, use the example below, replacing the placeholders surrounded by <> with your values:

1import requests
2
3HEADERS = {"x-api-key": "<YOUR_API_KEY>", "Content-Type": "application/json"}
4BASE_URL = "https://api.twelvelabs.io/v1.3"
5
6response = requests.post(
7    f"{BASE_URL}/knowledge-stores",
8    headers=HEADERS,
9    json={"name": "<YOUR_KNOWLEDGE_STORE_NAME>"}
10)
11store_id = response.json()["_id"]
12print(f"Knowledge store created: {store_id}")

Ingestion config

Jockey works out of the box with default extraction. To shape what the platform emphasizes during indexing, pass an ingestion_config when you create the knowledge store:

1response = requests.post(
2    f"{BASE_URL}/knowledge-stores",
3    headers=HEADERS,
4    json={
5        "name": "Marketing Videos",
6        "ingestion_config": {
7            "enrichment_config": {
8                "description": "Focus on brand mentions, product appearances, and audience reactions"
9            }
10        }
11    }
12)

For guidance on choosing between default extraction, a natural-language description, and a JSON schema, see Configure ingestion.

Metadata

Attach key-value pairs to a knowledge store for your own tracking. Maximum 16 pairs with string keys and values.

1response = requests.post(
2    f"{BASE_URL}/knowledge-stores",
3    headers=HEADERS,
4    json={
5        "name": "Q1 Marketing Videos",
6        "metadata": {
7            "team": "marketing",
8            "quarter": "2026-Q1"
9        }
10    }
11)

Jupyter notebook

Download the notebook to run this guide interactively.

API reference