Documents — Indexes

A collection needs no migration; it appears on first write. Indexes are still worth declaring so queries stay fast and unique constraints are enforced.

Declaring indexes

Indexes live on the model as a tuple of dicts. The shape mirrors what Mongo’s create_index accepts:

app/models/article.py

class Article(Document):
    indexes = (
        {"keys": [("slug", 1)], "unique": True},
        {
            "keys": [("author_id", 1), ("created_at", -1)],
            "name": "author_recent",
        },
        {"keys": [("tags", 1)], "name": "tags_asc"},
    )

Key	Meaning
keys	Sequence of (field, direction) pairs — 1 ascending, -1 descending
unique	Reject duplicate values for that key
name	Explicit index name (Mongo otherwise invents one)
other options	Passed through to the store (TTL, partial filters, …) when the driver supports them

The memory store enforces unique in process so tests catch collisions without Mongo. Other options are best-effort there.

Syncing

terminal

smith documents:index
smith documents:index --model Article
smith documents:index --pretend
smith documents:show --database mongodb

From code:

app/http/controllers/example_controller.py

await Article.sync_indexes()

documents:show lists collections, document counts, and indexes for a named document connection. If more than one document connection exists, pass --database.

What indexes are not

Almasix does not ship a Schema Blueprint for collections. Declaring indexes on the model (and syncing them) is the supported path — the same honesty rule as skipping SQL migrations for documents. Text indexes, TTL indexes, and Atlas Search / vector index definitions can be passed as options where Mongo accepts them; first-class helpers for those index kinds are not shipped yet.