Django ORM Intellisense

Framework-aware autocomplete, hover, go-to-definition, and diagnostics for Django ORM — right inside VS Code.

Stop guessing lookup paths like author__profile__timezone__icontains or relation strings like "app_label.ModelName". This extension understands your Django project's models, relations, managers, and querysets at runtime, and gives you real-time feedback as you type.

Features

ORM Lookup Path Completion & Validation

Autocomplete for keyword arguments in filter(), exclude(), get(), order_by(), and other queryset methods. The extension resolves foreign key chains, reverse relations, and lookup operators (__icontains, __gte, etc.) based on your actual model graph.

# Autocomplete suggests: author, author__profile, author__profile__timezone, ...
Product.objects.filter(author__profile__timezone__icontains="Asia")

Invalid lookup paths are flagged with diagnostics so you catch typos before they hit production.

String Lookup Path Intelligence

Completion, hover, and go-to-definition for string-based ORM paths used in values(), values_list(), select_related(), prefetch_related(), only(), defer(), and more.

Product.objects.select_related("author__profile")
Product.objects.values("author__profile__timezone")

Relation String Resolution

Autocomplete and hover for Django relation-string targets inside ForeignKey(...), OneToOneField(...), and ManyToManyField(...) declarations.

Queryset Receiver Inference

The extension infers the base model for queryset variables, method return values, and chained calls:

products = Product.objects.active()
products.filter(...)  # knows this is a Product queryset

# Also works with:
# - self.get_queryset().filter(...)
# - cls.available().filter(...)
# - super().base_queryset().filter(...)
# - for item in qs: item.  (loop target inference)
# - build_products().filter(...)  (helper return inference)

Expression & Annotation Support

• Q(), F(), When(), Case() expression field paths
• annotate() / alias() member propagation to downstream lookups and instance access
• Subquery, OuterRef, aggregate and window expression field paths
• Meta index and constraint field contexts

Multiline Support

Handles multiline parenthesized expressions, chained method calls, and complex nested Q/When/Case constructs.

Import & Re-export Resolution

Hover on from package import Symbol shows the origin module behind __init__.py re-exports. Go-to-definition navigates to the actual source, not the re-export shim.

Django Builtin Method Hover

Hover on Django builtin methods like .save(), .delete(), .filter(), .get() shows method signatures with documentation.

Pylance Integration

Optional suppression of Pylance false positives for dynamic Django ORM attributes that cannot be inferred statically.

Requirements

• VS Code 1.90.0 or later
• Python 3.10+ with Django installed in the project's virtual environment
• The extension runs a Python analysis daemon that needs access to your Django project's dependencies

Quick Start

1. Install the extension from the VS Code Marketplace.
2. Open a Django project in VS Code.
3. Run Django ORM: Select Python Interpreter (Cmd+Shift+P) and choose the project's virtualenv.
4. The extension auto-detects DJANGO_SETTINGS_MODULE from manage.py. For multi-settings projects, run Django ORM: Select Settings Module.
5. Start typing in a filter(), values(), or relation field — completions appear automatically.

Commands

Configuration

Architecture

The extension uses a split architecture:

• TypeScript client — VS Code integration, UI, provider registration
• Python daemon — Semantic analysis via hybrid static indexing + runtime Django inspection

The daemon bootstraps django.setup() inside your project's virtualenv to access the full model registry, then builds a semantic graph of models, fields, relations, managers, and querysets. Static indexing handles import resolution and re-export chains without Django being importable.

Known Limitations

• The daemon must run inside the same Python environment where Django and project dependencies are installed.
• Very large projects (1000+ models) may experience brief delays on first indexing; subsequent operations use cached indexes.
• Anonymous or dynamic model classes (e.g., generated at runtime by meta-programming) may not be fully indexed.

License

MIT