Backend Developer Guide

Learn how to create type-safe RPC handlers using Pydantic models and the @websocket_rpc decorator.

Creating Your First Handler

1. Define Pydantic Models

Create request and response models using Pydantic v2:


# core/centrifugo_handlers.py
from pydantic import BaseModel, Field
from typing import List, Optional
 
class TaskStatsParams(BaseModel):
    """Request parameters for task statistics."""
    user_id: str = Field(..., description="User ID to fetch stats for")
    include_completed: bool = Field(True, description="Include completed tasks")
 
class TaskStatsResult(BaseModel):
    """Task statistics response."""
    total: int = Field(..., description="Total number of tasks")
    completed: int = Field(..., description="Number of completed tasks")
    pending: int = Field(..., description="Number of pending tasks")
    user_id: str = Field(..., description="User ID")

2. Create RPC Handler

Use the @websocket_rpc decorator to register your handler:


from django_cfg.apps.integrations.centrifugo.decorators import websocket_rpc
 
@websocket_rpc("tasks.get_stats")
async def get_task_stats(conn, params: TaskStatsParams) -> TaskStatsResult:
    """
    Get task statistics for a user.
 
    This handler retrieves comprehensive task statistics including
    total, completed, and pending task counts.
    """
    from apps.tasks.models import Task
 
    # Use params as Pydantic model
    user_id = params.user_id
 
    # Query database
    total = Task.objects.filter(user_id=user_id).count()
    completed = Task.objects.filter(user_id=user_id, status='completed').count()
    pending = total - completed
 
    # Return Pydantic model
    return TaskStatsResult(
        total=total,
        completed=completed,
        pending=pending,
        user_id=user_id
    )

3. Register Handlers

Import handlers in your AppConfig.ready():


# core/apps.py
from django.apps import AppConfig
 
class CoreConfig(AppConfig):
    name = "core"
    default_auto_field = "django.db.models.BigAutoField"
 
    def ready(self):
        """Import handlers to register them with the router."""
        from . import centrifugo_handlers

Handler Best Practices

Use Type Hints

Type hints are required - they drive code generation:


# ✅ GOOD - Pydantic models with type hints
@websocket_rpc("user.get")
async def get_user(conn, params: GetUserParams) -> UserResult:
    ...
 
# ❌ BAD - No type hints
@websocket_rpc("user.get")
async def get_user(conn, params):
    ...

Add Field Descriptions

Descriptions appear in generated client documentation:


class UserParams(BaseModel):
    user_id: str = Field(..., description="Unique user identifier")
    include_profile: bool = Field(
        True,
        description="Include full profile data in response"
    )

Generates TypeScript documentation:


interface UserParams {
  /** Unique user identifier */
  user_id: string;
  /** Include full profile data in response */
  include_profile?: boolean;
}

Write Handler Docstrings

Docstrings appear in README files:


@websocket_rpc("tasks.create")
async def create_task(conn, params: CreateTaskParams) -> TaskResult:
    """
    Create a new task.
 
    Creates a task for the specified user with the given title and description.
    Automatically sets created_at timestamp and initializes status to 'pending'.
 
    Returns the created task with its generated ID.
    """
    ...

Use Async Handlers

All handlers must be async:


# ✅ GOOD
@websocket_rpc("tasks.list")
async def list_tasks(conn, params: ListParams) -> ListResult:
    tasks = await Task.objects.filter(user_id=params.user_id).alist()
    return ListResult(tasks=[...])
 
# ❌ BAD - sync handler
@websocket_rpc("tasks.list")
def list_tasks(conn, params: ListParams) -> ListResult:
    ...

Naming Conventions

Method Names

Use dot notation for namespacing:


@websocket_rpc("tasks.list")       # List tasks
@websocket_rpc("tasks.create")     # Create task
@websocket_rpc("tasks.update")     # Update task
@websocket_rpc("tasks.delete")     # Delete task
 
@websocket_rpc("users.get")        # Get user
@websocket_rpc("users.profile")    # Get user profile
 
@websocket_rpc("system.health")    # System health check
@websocket_rpc("system.stats")     # System statistics

This generates organized client methods:


# Python
api.tasks_list(...)
api.tasks_create(...)
api.users_get(...)
api.system_health(...)


// TypeScript
api.tasksList(...)
api.tasksCreate(...)
api.usersGet(...)
api.systemHealth(...)


// Go
api.TasksList(ctx, ...)
api.TasksCreate(ctx, ...)
api.UsersGet(ctx, ...)
api.SystemHealth(ctx, ...)

Model Names

Follow these conventions:


# Request models: <Action><Entity>Params
class CreateTaskParams(BaseModel): ...
class UpdateUserParams(BaseModel): ...
class ListTasksParams(BaseModel): ...
 
# Response models: <Entity>Result or <Action><Entity>Result
class TaskResult(BaseModel): ...
class UserResult(BaseModel): ...
class TaskListResult(BaseModel): ...

Advanced Patterns

Optional Parameters

Use Optional for optional fields:


from typing import Optional
 
class SearchParams(BaseModel):
    query: str = Field(..., description="Search query")
    limit: Optional[int] = Field(None, description="Max results")
    offset: Optional[int] = Field(None, description="Result offset")

Generates:


interface SearchParams {
  query: string;
  limit?: number | null;
  offset?: number | null;
}

Lists and Nested Models

Use List for arrays and nest models:


from typing import List
 
class Tag(BaseModel):
    name: str
    color: str
 
class Task(BaseModel):
    id: int
    title: str
    tags: List[Tag]
 
class TaskListResult(BaseModel):
    tasks: List[Task]
    total: int

Generates:


interface Tag {
  name: string;
  color: string;
}
 
interface Task {
  id: number;
  title: string;
  tags: Tag[];
}
 
interface TaskListResult {
  tasks: Task[];
  total: number;
}

Using Connection Object

The conn parameter provides access to connection metadata:


@websocket_rpc("user.profile")
async def get_profile(conn, params: ProfileParams) -> ProfileResult:
    # Access user from connection
    user_id = conn.user_id  # From JWT token
 
    # Use in business logic
    profile = await UserProfile.objects.get(user_id=user_id)
 
    return ProfileResult(
        username=profile.username,
        email=profile.email
    )

Error Handling

Raise exceptions for error cases:


from django.core.exceptions import ObjectDoesNotExist
 
@websocket_rpc("tasks.get")
async def get_task(conn, params: GetTaskParams) -> TaskResult:
    try:
        task = await Task.objects.aget(id=params.task_id)
        return TaskResult(...)
    except ObjectDoesNotExist:
        raise ValueError(f"Task {params.task_id} not found")

Clients receive error responses:


try:
    result = await api.tasks_get(GetTaskParams(task_id=999))
except Exception as e:
    print(f"Error: {e}")  # "Task 999 not found"

Database Queries

Use Django ORM async methods:


@websocket_rpc("tasks.list")
async def list_tasks(conn, params: ListParams) -> ListResult:
    # Async query
    tasks = await Task.objects.filter(
        user_id=params.user_id
    ).order_by('-created_at').alist()
 
    # Transform to Pydantic
    task_list = [
        TaskItem(id=t.id, title=t.title, status=t.status)
        for t in tasks
    ]
 
    return ListResult(tasks=task_list, total=len(task_list))

Testing Handlers

Unit Tests

Test handlers directly:


# core/tests/test_centrifugo_handlers.py
import pytest
from core.centrifugo_handlers import get_task_stats, TaskStatsParams
 
@pytest.mark.asyncio
async def test_get_task_stats():
    # Create test data
    user = await User.objects.acreate(username="test")
    await Task.objects.acreate(user=user, status="completed")
    await Task.objects.acreate(user=user, status="pending")
 
    # Call handler
    result = await get_task_stats(
        conn=None,  # Mock connection
        params=TaskStatsParams(user_id=user.id)
    )
 
    # Assert results
    assert result.total == 2
    assert result.completed == 1
    assert result.pending == 1

Integration Tests

Test with generated clients:


@pytest.mark.asyncio
async def test_rpc_client_integration():
    from opensdk.python import CentrifugoRPCClient, APIClient
 
    # Setup client
    rpc = CentrifugoRPCClient(
        url=settings.CENTRIFUGO_URL,
        token=generate_test_token(),
        user_id="test-user"
    )
    await rpc.connect()
 
    api = APIClient(rpc)
 
    # Call RPC method
    result = await api.tasks_get_stats(TaskStatsParams(user_id="test-user"))
 
    # Verify
    assert result.total >= 0
    assert result.completed >= 0

Complete Example

Here’s a complete CRUD example for a Todo app:


# core/centrifugo_handlers.py
from pydantic import BaseModel, Field
from typing import List, Optional
from django_cfg.apps.integrations.centrifugo.decorators import websocket_rpc
from apps.todos.models import Todo
 
# ========================================
# Models
# ========================================
 
class TodoItem(BaseModel):
    id: int
    title: str
    completed: bool
    created_at: str
 
class CreateTodoParams(BaseModel):
    title: str = Field(..., description="Todo title")
 
class UpdateTodoParams(BaseModel):
    id: int = Field(..., description="Todo ID")
    title: Optional[str] = Field(None, description="New title")
    completed: Optional[bool] = Field(None, description="Completion status")
 
class DeleteTodoParams(BaseModel):
    id: int = Field(..., description="Todo ID to delete")
 
class ListTodosParams(BaseModel):
    completed: Optional[bool] = Field(None, description="Filter by completion")
    limit: Optional[int] = Field(10, description="Max results")
 
class TodoResult(BaseModel):
    todo: TodoItem
 
class TodoListResult(BaseModel):
    todos: List[TodoItem]
    total: int
 
class SuccessResult(BaseModel):
    success: bool
    message: str
 
# ========================================
# Handlers
# ========================================
 
@websocket_rpc("todos.create")
async def create_todo(conn, params: CreateTodoParams) -> TodoResult:
    """Create a new todo item."""
    todo = await Todo.objects.acreate(
        user_id=conn.user_id,
        title=params.title,
        completed=False
    )
 
    return TodoResult(
        todo=TodoItem(
            id=todo.id,
            title=todo.title,
            completed=todo.completed,
            created_at=todo.created_at.isoformat()
        )
    )
 
@websocket_rpc("todos.update")
async def update_todo(conn, params: UpdateTodoParams) -> TodoResult:
    """Update an existing todo item."""
    todo = await Todo.objects.aget(id=params.id, user_id=conn.user_id)
 
    if params.title is not None:
        todo.title = params.title
    if params.completed is not None:
        todo.completed = params.completed
 
    await todo.asave()
 
    return TodoResult(
        todo=TodoItem(
            id=todo.id,
            title=todo.title,
            completed=todo.completed,
            created_at=todo.created_at.isoformat()
        )
    )
 
@websocket_rpc("todos.delete")
async def delete_todo(conn, params: DeleteTodoParams) -> SuccessResult:
    """Delete a todo item."""
    await Todo.objects.filter(id=params.id, user_id=conn.user_id).adelete()
 
    return SuccessResult(
        success=True,
        message=f"Todo {params.id} deleted"
    )
 
@websocket_rpc("todos.list")
async def list_todos(conn, params: ListTodosParams) -> TodoListResult:
    """List todos with optional filtering."""
    query = Todo.objects.filter(user_id=conn.user_id)
 
    if params.completed is not None:
        query = query.filter(completed=params.completed)
 
    query = query.order_by('-created_at')[:params.limit]
    todos = await query.alist()
 
    todo_items = [
        TodoItem(
            id=t.id,
            title=t.title,
            completed=t.completed,
            created_at=t.created_at.isoformat()
        )
        for t in todos
    ]
 
    return TodoListResult(
        todos=todo_items,
        total=len(todo_items)
    )

Next Steps

Client Generation - Generate type-safe clients
Frontend Guide - Use clients in your frontend
API Reference - Complete API documentation

Handler Checklist Before generating clients:

✅ All handlers use Pydantic models
✅ Type hints on all parameters and returns
✅ Descriptive field descriptions
✅ Handler docstrings written
✅ Handlers are async
✅ Handlers registered in apps.py