Models

Module: promptframe.models

Pydantic data models that represent the contents of loaded YAML prompt files.

---

Metadata

class Metadata(BaseModel)

Represents the metadata block at the top of every prompt YAML file.

Field	Type	Required	Description
type	"prompt" | "model_prompt"	✅	File type discriminator
name	str	✅	Human-readable name for this collection
description	str | None	—	Short description
tags	List[str] | None	—	Free-form tags
project	str | None	—	Project identifier

---

Prompt

class Prompt(BaseModel)

A single prompt entry inside a type: prompt YAML file.

Field	Type	Required	Description
pid	str	✅	Unique prompt identifier
description	str | None	—	What this prompt does
input_variables	List[str] | None	—	Named {placeholder} variables
prompt	str	✅	The prompt text, may contain {placeholders}

Methods

format(**kwargs) -> str

Return the prompt text with {placeholders} filled from keyword arguments.

prompt = Prompt(pid="greet", prompt="Hello, {name}!")
prompt.format(name="Alice")  # "Hello, Alice!"

__str__() -> str

Returns the raw prompt text.

---

PromptDataModel

class PromptDataModel(BaseModel)

A single prompt entry inside a type: model_prompt YAML file. Used to inject per-field instructions into LLMBaseModel schemas.

Field	Type	Required	Description
pid	str	✅	Unique prompt identifier
description	str | None	—	Description of this prompt's purpose
input_variables	List[str] | None	—	Named {placeholder} variables
model_attribute_id	str	✅	Key used to bind this prompt to an LLMField on a model
input_instruction	str | None	—	Instruction injected into input schemas
output_instruction	str | None	—	Instruction injected into output schemas

Validation: At least one of input_instruction or output_instruction must be provided, or a ValueError is raised.

---

PromptYAML

class PromptYAML(PromptYAMLBase)

The fully parsed representation of a type: prompt YAML file. Returned by PromptRegistry.load_prompt().

Field	Type	Description
version	float	Schema version
metadata	Metadata	File metadata
prompts	List[Prompt]	All prompts in the file

Properties

prompt_dict -> Dict[str, Prompt]

Cached dictionary keyed by pid. Built once on first access.

Methods

get(pid: str) -> Prompt | None

Return a Prompt by its pid, or None if not found.

p = prompts.get("summarize_text")

__getattr__(name: str) -> Prompt

Attribute-style access to prompts by pid. Raises PromptNotFoundError if the pid does not exist.

p = prompts.summarize_text       # equivalent to prompts.get("summarize_text")
p = prompts.prompt_dict["summarize_text"]  # dict access

---

PromptDataModelYAML

class PromptDataModelYAML(PromptYAMLBase)

The fully parsed representation of a type: model_prompt YAML file. Returned by PromptRegistry.load_model_prompt().

Field	Type	Description
version	float	Schema version
metadata	Metadata	File metadata
prompts	List[PromptDataModel]	All model prompts in the file

Properties

prompt_dict -> Dict[str, PromptDataModel]

Cached dictionary keyed by pid.

prompt_model_dict -> Dict[str, PromptDataModel]

Cached dictionary keyed by model_attribute_id. This is the mapping you pass to LLMBaseModel methods.

mp = registry.load_model_prompt("field_prompts")
MyModel.get_format_instructions_with_prompt(
    prompt_model_dict=mp.prompt_model_dict
)

Methods

get(pid: str) -> PromptDataModel | None

Return a PromptDataModel by its pid, or None if not found.

__getattr__(name: str) -> PromptDataModel

Attribute-style access to prompts by pid. Raises PromptNotFoundError if the pid does not exist.