Adapt Django¶

Adapt Django — mvp.adapt_django

Cluster: Code Intelligence | Type: component | MCP Tools: 44

Overview¶

Django code-generation block that produces complete project scaffolds, apps, models, views, URLs, admin registrations, templates, forms, serializers, viewsets, migrations, and test files — all as in-memory strings with no runtime Django dependency. Supports natural-language-to-model synthesis via nl_to_models, static code validation, and optional file writing to disk when base_dir is set. Design tokens can be injected as CSS variables to align generated templates with a design system.

When to use:

• Scaffolding a Django project or app from a structured model and view specification
• Converting a natural-language description of a data model into Django model code
• Generating REST API viewsets and serializers as part of an automated backend-generation pipeline
• Producing migration files and settings configurations for different databases

Example:

from mvp.adapt_django import AdaptDjangoBlock, DjangoInput, ModelSpec, FieldSpec

block = AdaptDjangoBlock(name="django")
result = block.infer(DjangoInput(
    operation="generate_models",
    app_name="blog",
    models=[ModelSpec(
        name="Post",
        fields=[FieldSpec(name="title", field_type="CharField", max_length=200),
                FieldSpec(name="body", field_type="TextField")],
        str_field="title",
    )],
))
# result.value.files → list of GeneratedFile with path and content

Works well with: adapt_webpage, meta_programming, adapt_ui_design

Public API¶

DjangoModelRecommendation¶

Validated Django data-model recommendation decision record.

Methods:

model_count() -> int¶

field_count() -> int¶

to_metadata() -> dict[str, Any]¶

LLMDjangoRuntime¶

Provider-neutral django-model recommendation runtime over G6's LLM caller.

Constructor:

Methods:

synthesize(nl_description: str) -> Result[list[ModelSpec]]¶

DjangoPlanner¶

Runtime-first facade with the deterministic floor as honest fallback.

Constructor:

Methods:

recommend_models(nl_description: str, app_name: str = '') -> DjangoModelRecommendation¶

AdaptDjangoBlock(AIBlock[DjangoInput, DjangoOutput, dict])¶

Django project and app code generator.

Methods:

infer(data: DjangoInput) -> Result[DjangoOutput]¶

NLModelSynthesizer¶

Convert a natural-language description into a list of Django ModelSpecs.

Constructor:

Methods:

synthesize(nl_description: str) -> Result[list[ModelSpec]]¶

RunnerResult¶

Result of a runner operation.

DjangoProjectRunner¶

Runs subprocess operations against a generated Django project.

Constructor:

Methods:

apply_migrations(timeout: int = 120) -> RunnerResult¶

Run manage.py migrate --no-input in the project directory.

validate_project(deploy: bool = True, timeout: int = 30) -> RunnerResult¶

Run manage.py check and optionally manage.py check --deploy.

FieldSpec(BaseModel)¶

Specification for a single Django model field.

ModelSpec(BaseModel)¶

Specification for a Django model class.

ViewSpec(BaseModel)¶

Specification for a Django class-based view.

GeneratedFile(BaseModel)¶

A single generated file with path and content.

DjangoInput(BaseModel)¶

Input to AdaptDjangoBlock.

DjangoOutput(BaseModel)¶

Output from AdaptDjangoBlock.

AdaptDjangoMCPBlock(AIBlock[MCPDjangoInput, MCPDjangoOutput, dict])¶

31-op Django code-generation block with SQLite persistence.

Methods:

infer(data: MCPDjangoInput) -> Result[MCPDjangoOutput]¶

close() -> None¶

Close the underlying store connection.

MCPDjangoInput(BaseModel)¶

Input for the MCP Django block (31-op schema, 27 exposed as MCP tools).

MCPDjangoOutput(BaseModel)¶

Output from the MCP Django block.

Methods:

model_post_init(__context: Any) -> None¶

DjangoStore¶

Sync SQLite store with 7 tables.

Constructor:

Methods:

save_project(name: str, style: str = 'monolith', database: str = 'sqlite3', auth_backend: str = 'django', config_json: str = '{}') -> str¶

get_project(name: str) -> dict[str, Any] | None¶

list_projects() -> list[dict[str, Any]]¶

delete_project(name: str) -> bool¶

save_app(project_name: str, name: str, config_json: str = '{}') -> str¶

list_apps(project_name: str = '') -> list[dict[str, Any]]¶

save_model(project_name: str, app_name: str, name: str, fields_json: str = '[]', meta_ordering_json: str = '[]', str_field: str = '', abstract: bool = False) -> str¶

get_model(app_name: str, name: str) -> dict[str, Any] | None¶

list_models(app_name: str = '') -> list[dict[str, Any]]¶

delete_model(app_name: str, name: str) -> bool¶

save_view(project_name: str, app_name: str, model_name: str, view_type: str, name: str, config_json: str = '{}') -> str¶

list_views(app_name: str = '') -> list[dict[str, Any]]¶

save_template(project_name: str, app_name: str, path: str, content: str, file_type: str = 'html') -> str¶

list_templates(app_name: str = '') -> list[dict[str, Any]]¶

save_api(project_name: str, app_name: str, api_type: str, model_name: str, config_json: str = '{}') -> str¶

list_apis(app_name: str = '') -> list[dict[str, Any]]¶

count_all() -> dict[str, int]¶

save_written_project(project_name: str, base_dir: str, manifest: list[str]) -> str¶

get_written_project(project_name: str) -> dict[str, Any] | None¶

list_written_projects() -> list[dict[str, Any]]¶

text_search(query: str, top_k: int = 5) -> list[dict[str, Any]]¶

close() -> None¶

Close the database connection.

Functions¶

agentic_planner_enabled(default_enabled: bool = True) -> bool¶

Decide whether the agentic django-model planner should be used.

recommend_models_floor(nl_description: str, app_name: str = '') -> list[ModelSpec]¶

Deterministic regex/heuristic NL->models selector (the demoted floor).

MCP Tools¶