Record Factories, Seeders, Fixtures & Pydantic Bridge

Internal engineering reference for the Sillo ORM factory system, seeder, fixture loader, Pydantic model generation, and database exception handlers.

Source: core/sillo/record/factories.py, core/sillo/record/helpers.py, core/sillo/record/pydantic.py, core/sillo/record/exceptions.py

---

1. Overview

The factory and seeding layer provides tools for populating the database with test data, loading fixtures, and bridging between Tortoise models and Pydantic schemas. It also includes exception handlers that convert database errors into proper HTTP responses.

graph TB
    subgraph "Factory System"
        A[Factory]
        B[FactoryBuilder]
    end
    subgraph "Seeding"
        C[Seeder]
        D[FixtureLoader]
    end
    subgraph "Pydantic Bridge"
        E[pydantic_model_from_tortoise]
    end
    subgraph "Exception Handlers"
        F[register_db_exception_handlers]
        G[handle_does_not_exist]
        H[handle_integrity_error]
        I[handle_validation_error]
        J[handle_operational_error]
    end
    subgraph "Database"
        K[Model]
        L[Tortoise ORM]
    end

    A -->|make/create| K
    B -->|registry| A
    C -->|seed| K
    D -->|load_all| K
    E -->|generates| Pydantic
    F --> G
    F --> H
    F --> I
    F --> J

---

2. Factory

File: core/sillo/record/factories.py

2.1 Base Class

class Factory:
    model: type | None = None
    definition: Callable[[], dict[str, Any]] = dict

    @classmethod
    def make(cls, overrides=None) -> Any:
        data = {**cls.definition(), **(overrides or {})}
        return cls.model(**data)

    @classmethod
    async def create(cls, overrides=None) -> Any:
        instance = cls.make(overrides)
        await instance.save()
        return instance

    @classmethod
    async def create_many(cls, count, overrides=None) -> list[Any]:
        instances = []
        for _ in range(count):
            instance = await cls.create(overrides)
            instances.append(instance)
        return instances

    @classmethod
    def state(cls, **kwargs) -> Callable:
        def modifier():
            return {**cls.definition(), **kwargs}
        return modifier

2.2 Defining a Factory

from uuid import uuid4
from sillo.record import Factory

class UserFactory(Factory):
    model = User
    definition = lambda: {
        "email": f"user{uuid4().hex[:8]}@test.com",
        "name": "Test User",
        "is_active": True,
    }

Requirements:

• model: the Tortoise model class.
• definition: a callable returning a dict of default attributes. Each call should produce unique values (e.g. UUIDs) to avoid constraint violations.

2.3 make: Unsaved Instance

@classmethod
def make(cls, overrides=None) -> Any:
    data = {**cls.definition(), **(overrides or {})}
    return cls.model(**data)

• Creates a model instance without saving to the database.
• overrides are merged on top of the definition.
• Useful for testing validation logic without hitting the DB.

user = UserFactory.make()
assert user.email.endswith("@test.com")
assert not user._saved_in_db

2.4 create: Saved Instance

@classmethod
async def create(cls, overrides=None) -> Any:
    instance = cls.make(overrides)
    await instance.save()
    return instance

• Creates and persists a model instance.
• Returns the saved instance with id populated.

user = await UserFactory.create()
assert user.id is not None
assert user._saved_in_db

2.5 create_many: Batch Creation

@classmethod
async def create_many(cls, count, overrides=None) -> list[Any]:
    instances = []
    for _ in range(count):
        instance = await cls.create(overrides)
        instances.append(instance)
    return instances

• Creates count instances, each with its own definition call (unique values).
• Note: This is N individual INSERTs, not a bulk insert. For large batches, consider using Model.bulk_create() directly.

users = await UserFactory.create_many(10)
assert len(users) == 10
assert len(set(u.email for u in users)) == 10  # all unique

2.6 state: Modifier Function

@classmethod
def state(cls, **kwargs) -> Callable:
    def modifier():
        return {**cls.definition(), **kwargs}
    return modifier

Returns a new definition function that merges overrides on top of the base definition. Used to create named variants:

admin_state = UserFactory.state(is_admin=True, role="admin")

class AdminFactory(Factory):
    model = User
    definition = admin_state

admin = await AdminFactory.create()
assert admin.is_admin is True

2.7 Factory Usage Diagram

sequenceDiagram
    participant Dev as Developer
    participant Factory as UserFactory
    participant Model as User Model
    participant DB as Database

    Note over Dev,DB: make(): no DB
    Dev->>Factory: UserFactory.make()
    Factory->>Factory: definition() → {email, name}
    Factory->>Model: User(**data)
    Factory-->>Dev: unsaved instance

    Note over Dev,DB: create(): with DB
    Dev->>Factory: UserFactory.create()
    Factory->>Factory: make() → instance
    Factory->>Model: instance.save()
    Model->>DB: INSERT INTO users
    DB-->>Model: id
    Factory-->>Dev: saved instance

    Note over Dev,DB: create_many(5)
    Dev->>Factory: UserFactory.create_many(5)
    loop 5 times
        Factory->>Factory: create()
        Factory->>DB: INSERT
    end
    Factory-->>Dev: list of 5 instances

---

3. FactoryBuilder

File: core/sillo/record/factories.py

3.1 Registry Pattern

class FactoryBuilder:
    def __init__(self):
        self._factories: dict[str, type[Factory]] = {}

    def register(self, name: str, factory: type[Factory]) -> None:
        self._factories[name] = factory

    def get(self, name: str) -> type[Factory]:
        if name not in self._factories:
            raise KeyError(f"Factory '{name}' not registered")
        return self._factories[name]

3.2 Usage

builder = FactoryBuilder()
builder.register("user", UserFactory)
builder.register("post", PostFactory)

# Later, by name:
user = await builder.get("user").create()
post = await builder.get("post").create(overrides={"author_id": user.id})

3.3 Design Rationale

The FactoryBuilder exists for scenarios where factories are registered dynamically (e.g., plugins, test fixtures loaded by name). For most projects, calling UserFactory.create() directly is simpler.

---

4. Seeder

File: core/sillo/record/helpers.py (class Seeder)

4.1 Class Definition

class Seeder:
    def __init__(self, db_manager):
        self._db = db_manager
        self._records: list[tuple[type, dict[str, Any]]] = []

    def seed(self, model, records) -> Seeder:
        for record in records:
            self._records.append((model, record))
        return self

    async def run(self, *, batch_size: int = 100) -> int:
        count = 0
        for model, data in self._records:
            await model.create(**data)
            count += 1
        return count

4.2 Usage

from sillo.record import Seeder

seeder = Seeder(db_manager)
seeder.seed(User, [
    {"email": "alice@ex.com", "name": "Alice"},
    {"email": "bob@ex.com", "name": "Bob"},
])
seeder.seed(Post, [
    {"title": "Hello World", "author_id": 1},
])
count = await seeder.run()
print(f"Seeded {count} records")

4.3 Design Decisions

• Deferred execution: seed() only registers records; run() executes them. This allows chaining multiple seed() calls before running.
• Sequential insertion: Each record is created individually with model.create(). This is deliberate: it triggers model lifecycle events (e.g. before_create, after_create) and validation.
• Return type: run() returns the total count of records created.
• batch_size parameter: Present in the signature but not currently used for batching: each record is created individually. Reserved for future optimization.

4.4 Seeder Flow

sequenceDiagram
    participant Dev as Developer
    participant Seeder as Seeder
    participant Model as User.create()
    participant DB as Database

    Dev->>Seeder: Seeder(db)
    Dev->>Seeder: seed(User, [{...}, {...}])
    Note over Seeder: registers 2 records
    Dev->>Seeder: seed(Post, [{...}])
    Note over Seeder: registers 1 record

    Dev->>Seeder: run()
    loop For each registered record
        Seeder->>Model: model.create(**data)
        Model->>DB: INSERT
        DB-->>Model: id
    end
    Seeder-->>Dev: 3 (total count)

---

5. FixtureLoader

File: core/sillo/record/helpers.py (class FixtureLoader)

5.1 Directory Convention

fixtures/
  01_users.json      →  [{"email": "...", ...}, ...]
  02_posts.json      →  [{"title": "...", ...}, ...]
  tags.jsonl          →  {"name": "python"}\n{"name": "rust"}\n...

• Files are loaded in sorted order, so numeric prefixes control ordering when fixtures reference each other.
• Supported extensions: .json, .jsonl.
• All other files are ignored.

5.2 Class Definition

class FixtureLoader:
    SUFFIXES = (".json", ".jsonl")

    def __init__(self, directory: str, *, models: dict[str, Any] | None = None):
        self._dir = Path(directory)
        self._models: dict[str, Any] = dict(models or {})

5.3 load_all

async def load_all(self) -> int:
    count = 0
    for file_path in sorted(self._dir.glob("*")):
        if file_path.suffix not in self.SUFFIXES or not file_path.is_file():
            continue
        count += await self._load_file(file_path)
    return count

• Iterates all files in sorted order.
• Skips non-matching suffixes and directories.
• Returns total rows inserted.

5.4 load: Single Fixture

async def load(self, name: str) -> int:
    for ext in self.SUFFIXES:
        path = self._dir / f"{name}{ext}"
        if path.exists():
            return await self._load_file(path)
    raise FileNotFoundError(f"Fixture '{name}' not found in {self._dir}")

• Loads a specific fixture by name (without extension).
• Tries each suffix in order (.json first, then .jsonl).
• Raises FileNotFoundError if no matching file exists.

5.5 _parse: File Parsing

def _parse(self, path: Path) -> list[dict[str, Any]]:
    content = path.read_text()
    if path.suffix == ".jsonl":
        records = [json.loads(line) for line in content.splitlines() if line.strip()]
    else:
        records = json.loads(content)
    if not isinstance(records, list):
        records = [records]
    return records

JSON format:

• Array of objects: [{"email": "..."}, {"email": "..."}]
• Single object: {"email": "..."} (treated as a one-row fixture)

JSONL format:

• One JSON object per line.
• Blank lines are skipped.
• Always produces a list.

5.6 _resolve_model: Model Resolution

def _resolve_model(self, name: str) -> Any:
    if name in self._models:
        return self._models[name]

    from tortoise import Tortoise

    registry = {
        model_name.lower(): model
        for app_models in Tortoise.apps.values()
        for model_name, model in app_models.items()
    }

    stem = name.lower()
    candidates = [stem]
    if stem.endswith("ies"):
        candidates.append(stem[:-3] + "y")
    if stem.endswith("es"):
        candidates.append(stem[:-2])
    if stem.endswith("s"):
        candidates.append(stem[:-1])

    for candidate in candidates:
        if candidate in registry:
            return registry[candidate]

    known = ", ".join(sorted(registry)) or "none — has Tortoise been initialised?"
    raise LookupError(
        f"Fixture '{name}' matches no registered model. Known models: {known}. "
        f"Pass FixtureLoader(..., models={{'{name}': YourModel}}) to map it explicitly."
    )

Resolution order:

1. Explicit mapping: if models dict was provided and contains the name.
2. Exact match: users → user (lowercase).
3. Pluralization heuristics:

Fixture name	Candidates generated	Matches model
users	users, user	User
categories	categories, categorie, categor, category	Category
statuses	statuses, statuse, statu, status	Status
aliases	aliases, aliase, alias	Alias
taxes	taxes, taxe, tax	Tax

4. Failure: raises LookupError with the list of known models.

5.7 _load_file: Transactional Loading

async def _load_file(self, path: Path) -> int:
    records = self._parse(path)
    if not records:
        return 0
    model = self._resolve_model(path.stem)
    from tortoise.transactions import in_transaction
    async with in_transaction():
        for record in records:
            await model.create(**record)
    return len(records)

Key detail: The entire file is loaded inside a transaction. If any row fails validation or violates a constraint, the entire file is rolled back, no half-populated tables.

5.8 FixtureLoader Flow Diagram

sequenceDiagram
    participant Dev as Developer
    participant Loader as FixtureLoader
    participant FS as Filesystem
    participant Parse as _parse
    participant Resolve as _resolve_model
    participant DB as Database

    Dev->>Loader: FixtureLoader("fixtures/")
    Dev->>Loader: load_all()
    Loader->>FS: sorted glob("*")

    loop For each .json/.jsonl file
        Loader->>FS: read file
        Loader->>Parse: _parse(path)
        Parse-->>Loader: list of dicts

        Loader->>Resolve: _resolve_model(stem)
        Resolve->>Resolve: check explicit mapping
        Resolve->>Resolve: check pluralization
        Resolve-->>Loader: model class

        Loader->>DB: BEGIN TRANSACTION
        loop For each record
            Loader->>DB: model.create(**record)
        end
        Loader->>DB: COMMIT
    end

    Loader-->>Dev: total rows inserted

---

6. Pydantic Model Generation

File: core/sillo/record/pydantic.py

6.1 pydantic_model_from_tortoise

def pydantic_model_from_tortoise(
    model_class: type,
    *,
    name: str = "",
    exclude: list[str] | None = None,
    include: list[str] | None = None,
    optional_fields: list[str] | None = None,
) -> type[BaseModel]:

6.2 Type Mapping

def _tortoise_to_python_type(field_obj) -> type:
    mapping = {
        f.IntField: int,
        f.SmallIntField: int,
        f.BigIntField: int,
        f.FloatField: float,
        f.DecimalField: float,
        f.BooleanField: bool,
        f.CharField: str,
        f.TextField: str,
        f.DatetimeField: str,
        f.DateField: str,
        f.TimeDeltaField: float,
        f.JSONField: dict,
    }
    for tf, pt in mapping.items():
        if isinstance(field_obj, tf):
            return pt
    return str

Note: DatetimeField maps to str (ISO 8601), not datetime. This is intentional. Pydantic models are used for request validation where datetimes arrive as strings from JSON payloads.

6.3 Field Processing

for field_name, field_obj in meta.fields_map.items():
    if exclude and field_name in exclude:
        continue
    if include and field_name not in include:
        continue

    py_type = _tortoise_to_python_type(field_obj)
    default = None if field_obj.null else ...
    is_required = not field_obj.null and not field_obj.pk

    if field_name in optional_fields or field_obj.null:
        py_type = Optional[py_type]
        default = None

    if is_required and not field_obj.null:
        fields[field_name] = (py_type, Field(...))
    else:
        fields[field_name] = (py_type, Field(default=default))

Decision matrix:

Condition	Type	Default
Required, non-null	T	...
Nullable	Optional[T]	None
In optional_fields	Optional[T]	None
Primary key	T	None

6.4 Model Creation

return create_model(
    name or f"{model_class.__name__}Schema",
    __base__=BaseModel,
    **fields
)

Uses Pydantic’s create_model to dynamically generate the model class.

6.5 Usage Patterns

Create schema (exclude auto-generated fields):

UserCreate = pydantic_model_from_tortoise(
    User,
    name="UserCreate",
    exclude=["id", "created_at", "updated_at", "deleted_at"],
)

Update schema (all fields optional):

UserUpdate = pydantic_model_from_tortoise(
    User,
    name="UserUpdate",
    exclude=["id", "created_at", "updated_at", "deleted_at"],
    optional_fields=["email", "name", "bio"],
)

Response schema (include specific fields):

UserResponse = pydantic_model_from_tortoise(
    User,
    name="UserResponse",
    include=["id", "email", "name", "created_at"],
)

6.6 Integration with Routes

from sillo.record.pydantic import pydantic_model_from_tortoise

UserCreate = pydantic_model_from_tortoise(User, name="UserCreate",
    exclude=["id", "created_at", "updated_at"])

@app.post("/users", request_model=UserCreate)
async def create_user(request, response):
    data = request.validated_data.model_dump()
    user = await User.create(**data)
    return response.json(user.to_dict(), status_code=201)

6.7 Pydantic Bridge Flow

sequenceDiagram
    participant Dev as Developer
    participant Bridge as pydantic_model_from_tortoise
    participant Tortoise as User._meta.fields_map
    participant Pydantic as create_model
    participant Route as POST /users
    participant DB as Database

    Dev->>Bridge: pydantic_model_from_tortoise(User, name="UserCreate")
    Bridge->>Tortoise: iterate fields_map
    loop For each field
        Bridge->>Bridge: _tortoise_to_python_type(field)
        Bridge->>Bridge: determine optionality
    end
    Bridge->>Pydantic: create_model("UserCreate", **fields)
    Pydantic-->>Dev: UserCreate class

    Note over Dev,DB: Request arrives
    Route->>Route: validate body against UserCreate
    Route->>DB: User.create(**validated_data)
    DB-->>Route: user instance
    Route-->>Dev: JSON response

---

7. Database Exception Handlers

File: core/sillo/record/exceptions.py

7.1 Overview

Four exception handlers convert Tortoise/database errors into proper HTTP responses:

Exception	Handler	HTTP Status	Error
DoesNotExist	handle_does_not_exist	404	Not Found
IntegrityError	handle_integrity_error	409	Conflict
ValidationError	handle_validation_error	422	Validation Error
OperationalError	handle_operational_error	503	Service Unavailable

7.2 handle_does_not_exist

async def handle_does_not_exist(request, response, exc: DoesNotExist):
    return response.json(
        {"error": "Not Found", "detail": str(exc)},
        status_code=404,
    )

Triggered when Model.get() finds no matching row.

7.3 handle_integrity_error

async def handle_integrity_error(request, response, exc: IntegrityError):
    return response.json(
        {"error": "Conflict", "detail": str(exc)},
        status_code=409,
    )

Triggered on:

• Unique constraint violations (duplicate email, etc.)
• Foreign key violations
• NOT NULL violations

7.4 handle_validation_error

async def handle_validation_error(request, response, exc: ValidationError):
    return response.json(
        {"error": "Validation Error", "detail": str(exc)},
        status_code=422,
    )

Triggered on Tortoise model validation failures (type mismatches, out-of-range values, etc.).

7.5 handle_operational_error

async def handle_operational_error(request, response, exc: OperationalError):
    return response.json(
        {"error": "Service Unavailable", "detail": "Database unavailable"},
        status_code=503,
    )

Triggered on:

• Connection refused
• Server timeout
• Network partition

Note: The detail message is generic (“Database unavailable”) to avoid leaking internal infrastructure details to the client.

7.6 register_db_exception_handlers

def register_db_exception_handlers(app) -> None:
    app.add_exception_handler(DoesNotExist, handle_does_not_exist)
    app.add_exception_handler(IntegrityError, handle_integrity_error)
    app.add_exception_handler(ValidationError, handle_validation_error)
    app.add_exception_handler(OperationalError, handle_operational_error)

One-call registration. Usage:

from sillo import SilloApp
from sillo.record import register_db_exception_handlers

app = SilloApp()
register_db_exception_handlers(app)

7.7 Exception Handler Flow

sequenceDiagram
    participant Client as HTTP Client
    participant App as SilloApp
    participant Handler as Route Handler
    participant DB as Database

    Client->>App: POST /users {"email": "dup@ex.com"}
    App->>Handler: create_user()
    Handler->>DB: User.create(email="dup@ex.com")
    DB-->>Handler: IntegrityError (duplicate key)

    Handler-->>App: raises IntegrityError
    App->>App: handle_integrity_error()
    App-->>Client: 409 {"error": "Conflict", "detail": "duplicate key..."}

---

8. Complete Factory + Seeder + Fixture Pattern

8.1 Project Structure

myapp/
  models.py          # Tortoise models
  factories.py       # Factory definitions
  fixtures/
    01_users.json
    02_posts.jsonl
  seeds.py           # Seeder script
  tests/
    conftest.py      # Factory fixtures
    test_users.py

8.2 Factory Definitions

myapp/factories.py

from uuid import uuid4
from sillo.record import Factory
from .models import User, Post

class UserFactory(Factory):
    model = User
    definition = lambda: {
        "email": f"user{uuid4().hex[:8]}@test.com",
        "name": "Test User",
        "is_active": True,
    }

class AdminFactory(Factory):
    model = User
    definition = UserFactory.state(is_admin=True, role="admin")

class PostFactory(Factory):
    model = Post
    definition = lambda: {
        "title": f"Post {uuid4().hex[:6]}",
        "body": "Lorem ipsum dolor sit amet.",
        "published": True,
    }

8.3 Seeder Script

myapp/seeds.py

import asyncio
from sillo.record import DatabaseConfig, DatabaseManager, Seeder
from myapp.models import User, Post

async def main():
    config = DatabaseConfig.postgres("myapp", "password")
    async with DatabaseManager(config).register_models("myapp.models") as db:
        seeder = Seeder(db)
        seeder.seed(User, [
            {"email": "admin@ex.com", "name": "Admin", "is_admin": True},
            {"email": "user@ex.com", "name": "User"},
        ])
        seeder.seed(Post, [
            {"title": "First Post", "body": "Hello World", "author_id": 1},
        ])
        count = await seeder.run()
        print(f"Seeded {count} records")

asyncio.run(main())

8.4 Fixture Files

fixtures/01_users.json

[
    {"email": "alice@ex.com", "name": "Alice", "is_active": true},
    {"email": "bob@ex.com", "name": "Bob", "is_active": true}
]

fixtures/02_posts.jsonl

{"title": "Alice's Post", "body": "Hello from Alice", "author_id": 1}
{"title": "Bob's Post", "body": "Hello from Bob", "author_id": 2}

8.5 Loading Fixtures

from sillo.record import FixtureLoader

loader = FixtureLoader("fixtures/", models={"posts": Post})
count = await loader.load_all()
print(f"Loaded {count} records")

8.6 Test Fixtures

tests/conftest.py

import pytest
from sillo.record import transaction, rollback

@pytest.fixture
async def user():
    async with transaction():
        u = await UserFactory.create()
        yield u
        await rollback()

@pytest.fixture
async def admin():
    async with transaction():
        u = await AdminFactory.create()
        yield u
        await rollback()

---

9. Integration with Exception Handlers

9.1 Full Application Setup

from sillo import SilloApp
from sillo.record import (
    setup_record, DatabaseConfig,
    register_db_exception_handlers,
)

app = SilloApp()

# Database
db = setup_record(
    app,
    DatabaseConfig.postgres("myapp", "password"),
    model_modules=["myapp.models"],
)

# Exception handlers
register_db_exception_handlers(app)

9.2 Error Response Format

All handlers return a consistent JSON structure:

{
    "error": "Human-readable error category",
    "detail": "Specific error message"
}

Status	Error	When
404	”Not Found”	Model.get() returns no rows
409	”Conflict”	Unique/FK constraint violation
422	”Validation Error”	Model validation failure
503	”Service Unavailable”	Database connection failure

---

10. Advanced Patterns

10.1 Factory with Related Models

class PostFactory(Factory):
    model = Post
    definition = lambda: {
        "title": f"Post {uuid4().hex[:6]}",
        "body": "Lorem ipsum.",
    }

# Usage with FK:
user = await UserFactory.create()
post = await PostFactory.create(overrides={"author_id": user.id})

10.2 Factory with Cast Fields

class EventFactory(Factory):
    model = Event
    definition = lambda: {
        "name": f"Event {uuid4().hex[:6]}",
        "metadata": {"location": "SF", "capacity": 100},
        "starts_at": datetime.now(timezone.utc).isoformat(),
    }

The cast system handles JSON encoding automatically on save.

10.3 Fixture with Explicit Model Mapping

When the fixture filename does not match the model name:

loader = FixtureLoader("fixtures/", models={
    "user_data": User,        # user_data.json → User model
    "blog_posts": Post,       # blog_posts.json → Post model
})

10.4 Selective Fixture Loading

loader = FixtureLoader("fixtures/")
await loader.load("01_users")  # only load users

10.5 Pydantic + Factory for Testing

from sillo.record.pydantic import pydantic_model_from_tortoise

UserCreate = pydantic_model_from_tortoise(User, exclude=["id", "created_at"])

def test_user_validation():
    data = UserCreate(email="test@example.com", name="Test")
    user = UserFactory.make(overrides=data.model_dump())
    assert user.email == "test@example.com"

---

11. Source File Reference

File	Contents
core/sillo/record/factories.py	Factory (make, create, create_many, state), FactoryBuilder
core/sillo/record/helpers.py	Seeder (seed, run), FixtureLoader (load_all, load, _parse, _resolve_model, _load_file), MigrationHelper (see doc 24)
core/sillo/record/pydantic.py	pydantic_model_from_tortoise, _tortoise_to_python_type
core/sillo/record/exceptions.py	handle_does_not_exist, handle_integrity_error, handle_validation_error, handle_operational_error, register_db_exception_handlers

---

12. Gotchas and Known Issues

1. create_many is N individual INSERTs: For large batches, use Model.bulk_create() directly for better performance.

2. definition must produce unique values: If the definition returns static values, create_many will fail on unique constraints.

3. FixtureLoader._resolve_model pluralization: The heuristic handles common English plurals (ies→y, es, s) but not irregular plurals (people→person, children→child). Use the explicit models mapping for those.

4. Fixture transaction scope. Each fixture file is loaded in its own transaction. If 02_posts.json references users from 01_users.json, both must succeed. A failure in posts does not roll back users.

5. Seeder.run() is sequential: Each record is created individually. For large seed datasets, consider Model.bulk_create() with pre-built instances.

6. Pydantic DatetimeField → str: This is intentional for request validation, but means the generated schema does not validate ISO 8601 format. Add a custom validator if needed.

7. Exception handler OperationalError hides details: The response says “Database unavailable” without specifics. This is correct for production but makes debugging harder. Check server logs for the actual error.

8. Factory.state returns a callable: It does not return a modified factory class. Assign the result to a variable or use it as the definition of a new factory.

9. Fixture model resolution requires Tortoise init. _resolve_model queries Tortoise.apps to find registered models. If Tortoise has not been initialized, the registry is empty and all lookups fail.

10. register_db_exception_handlers is idempotent: Calling it multiple times adds duplicate handlers. The last handler registered wins for each exception type, but the others still run (and produce no output since the response is already sent).