Lazy DevSearch hydration

Use the built-in DevSearchBackend when a local Django process needs a service-free search projection. This recipe enables per-index hydration on the first search, keeps later same-process manager changes visible, and makes the multi-term matching contract explicit.

Configure the disposable backend

Put the backend and setting in the same GENERAL_MANAGER mapping:

GENERAL_MANAGER = {
    "SEARCH_BACKEND": {
        "class": "general_manager.search.backends.dev.DevSearchBackend",
        "options": {},
    },
    "SEARCH_AUTO_REINDEX": True,
}

SEARCH_AUTO_REINDEX is opt-in and applies only to the selected DevSearchBackend; it is independent of DEBUG. The nested setting takes precedence over a top-level SEARCH_AUTO_REINDEX value. External search backends ignore this DevSearch-specific setting.

Declare an index and search it

Add a SearchConfig to the manager whose fields should be searchable:

from general_manager import GeneralManager, IndexConfig


class Project(GeneralManager):
    class SearchConfig:
        indexes = (
            IndexConfig(
                name="global",
                fields=("name", "status"),
                filters=("status",),
            ),
        )

With the setting enabled, the first search for global reindexes the configured Project managers in the serving process before evaluating the query:

from general_manager.search.backend_registry import get_search_backend


backend = get_search_backend()
result = backend.search("global", "alpha public", filters={"status": "public"})

for hit in result.hits:
    print(hit.id, hit.score)

Every distinct query term must match at least one indexed field token, so "alpha public" can match alpha in name and public in status, while a document missing either term is excluded. Repeating a term, as in "alpha alpha", does not increase its score.

Opt in a directly constructed backend

Direct construction stays inert by default. Enable hydration explicitly when a programmatic caller owns the backend instance:

from general_manager.search.backends.dev import DevSearchBackend


backend = DevSearchBackend(auto_reindex=True)
# Equivalent after construction:
# backend.configure_auto_reindex(True)
result = backend.search("global", "alpha")

Hydration is tracked independently per index. If manager discovery or reindexing raises, the original exception reaches the triggering search and a later search retries that index. After successful hydration, synchronous manager lifecycle updates in the same process upsert or delete documents as usual.

Keep the process boundary in mind

DevSearch is an in-memory, process-local projection. A separate python manage.py search_index --reindex process cannot populate the backend of an already running server, and writes made by another process are not observed. Restart the serving process so its next search can hydrate a fresh projection. Use a persistent external backend for production or cross-process search.

For the model and lifecycle details, see the search concept page and the search how-to. The generated signatures and compatibility notes are in the Search API reference.