A collection is where you store vectors (embeddings) and any metadata you want to filter or return in results. In DODIL, collections live in VBase (Milvus-backed) and are created using the DODIL Python SDK.

Create a collection (quick start)

Most vector apps start with:

• A primary key field (defaults to id)
• A vector field (defaults to vector)
• A known vector dimension (e.g., 768, 1024, 1536… depending on your embedding model)
• A metric (commonly COSINE)

Here’s a minimal create call:

from dodil import Client
from dodil.vbase import VBaseConfig
 
c = Client(service_account_id="...", service_account_secret="...")
 
vbase = c.vbase.connect(
    VBaseConfig(
        host="vbase-db-<id>.infra.dodil.cloud",
        port=443,
        scheme="https",
        db_name="db_<id>",
    )
)
 
collection_name = "docs_embeddings"
 
vbase.create_collection(
    collection_name=collection_name,
    dimension=1536,          # match your embedding model output
    metric_type="COSINE",   # COSINE, L2, IP
    auto_id=False,           # you control the id values
)
 
print("Created:", collection_name)
print("All collections:", vbase.list_collections())

What these parameters mean

• collection_name: A stable name for your app (e.g., docs_embeddings, support_tickets, product_catalog).
• dimension: The length of each embedding vector.
• metric_type:
  • COSINE → semantic similarity (most common for text embeddings)
  • L2 → Euclidean distance
  • IP → inner product
• auto_id:
  • False → you provide your own ids (recommended when you want deterministic control)
  • True → the database generates ids for you

Verify the collection

You can check whether a collection exists and inspect its settings.

exists = vbase.has_collection("docs_embeddings")
print("Exists?", exists)
 
if exists:
    info = vbase.describe_collection("docs_embeddings")
    print(info)

Next steps

Now that you have a collection, you can:

• Insert vectors (your embeddings + metadata)
• Build indexes for faster search
• Search using a query vector

Continue with:

• Insert vectors → /vbase/insert
• Search → /vbase/search
• Collections reference → /sdk/python/vbase