promptmeteo package#

Subpackages#

Submodules#

promptmeteo.api_formatter module#

class promptmeteo.api_formatter.APIFormatter(language: str, model_name: str, model_provider_name: str, api_version: str, api_protocol: str, api_style_instructions: List[str] = None, **kwargs)#

Bases: BaseUnsupervised

API Formatter Task.

This class initializes the API Formatter Task to correct and format APIs.

Parameters:

  • language (str) – Language for the API descriptions.

  • model_name (str) – Name of the model to be used for API generation.

  • model_provider_name (str) – Name of the model provider.

  • api_version (str) – Version of the API.

  • api_protocol (str) – Protocol of the API.

  • api_style_instructions (Optional[List[str]]) – Instructions for API style.

  • **kwargs (dict) – Additional keyword arguments.

  • Raises –

  • ------ –

  • ValueError – If api_protocol is not in the allowed protocols or if api_version is not in the correct format.

  • Example –

  • ------- –

  • APIFormatter (>>> from promptmeteo import) –

  • APIFormatter( (>>> model =) –

  • language="en" (...) –

:param : :param … api_version=”3.0.3”: :param : :param … api_protocol=”REST”: :param : :param … api_style_instructions=[“Use always camel case.”: :param “Do not use acronyms.”]: :param : :param … model_provider_name=”openai”: :param : :param … model_name=”gpt-3.5-turbo-16k”: :param : :param … model_provider_token=model_token: :param : :param … external_info={: :param … “servers”: :type … “servers”: [ :param … {: :param … “url”: :type … “url”: “http://localhost:8080/”, :param … “description”: :type … “description”: “Local environment”, :param … }: :param … ]: :param : :param … }: :param : :param … ): :param >>> model.train(api_code): :param >>> model.predict(api):

ALLOWED_PROTOCOLS = ['REST']#
TASK_TYPE: str = 'api-correction'#
classmethod load_model(model_path: str) → Self#

Load a saved model from disk.

model_pathstr

Path from where the model will be loaded.

Base

Loaded model instance.

Loads a model artifact to make new predictions.

model_pathstr

Path to the model artifact.

APIFormatter

Returns the loaded APIFormatter object.

predict(api_codes: List[str], external_info: dict) → List[str]#

Predict over new text samples.

examplesList[str]

List of text samples to predict.

List[str]

List of predictions.

Receive a list of API codes and return a list with the corrected APIs.

api_codesList[str]

List of API codes.

external_infodict

External information to be added to the APIs.

List[str]

List of corrected APIs.

train(api_codes: List[str]) → Self#
Trains the model given use cases and notes on behaviour. Check the

parameters and task behaviour in each specific model training docstring.

Train the APIFormatter to extract entities and parameters.

api_codesList[str]

List of API codes.

APIFormatter

Returns the trained APIFormatter object.

promptmeteo.api_generator module#

class promptmeteo.api_generator.APIGenerator(language: str, model_name: str, model_provider_name: str, api_version: str, api_protocol: str, api_style_instructions: List[str] | None, **kwargs)#

Bases: BaseSupervised

API Generator Task.

This class initializes the APIGenerator Task to create APIs based on descriptions.

Parameters:

  • language (str) – Language for the API descriptions.

  • model_name (str) – Name of the model to be used for API generation.

  • model_provider_name (str) – Name of the model provider.

  • api_version (str) – Version of the API.

  • api_protocol (str) – Protocol of the API.

  • api_style_instructions (Optional[List[str]]) – Instructions for API style.

  • **kwargs (dict) – Additional keyword arguments.

  • Raises –

  • ------ –

  • ValueError – If api_protocol is not in the allowed protocols or if api_version is not in the correct format.

  • Example –

  • ------- –

  • APIGenerator (>>> from promptmeteo import) –

  • APIGenerator( (>>> model =) –

  • language="en" (...) –

:param : :param … selector_k=5: :param : :param … api_version=”3.0.3”: :param : :param … api_protocol=”REST”: :param : :param … api_style_instructions=[“Use always camel case.”: :param “Do not use acronyms.”]: :param : :param … model_provider_name=”openai”: :param : :param … model_name=”gpt-3.5-turbo-16k”: :param : :param … model_provider_token=model_token: :param : :param … ): :param >>> model.train(api_description: :param api_code): :param >>> model.predict(“API for managing account access”):

ALLOWED_PROTOCOLS = ['REST']#
TASK_TYPE: str = 'api-generation'#
train(api_descriptions: List[str], api_codes: List[str]) → Self#
Trains the model given use cases and notes on behaviour. Check the

parameters and task behaviour in each specific model training docstring.

Train the APIGenerator from a list of APIs descriptions and contracts.

api_descriptionsList[str]

List of API descriptions.

api_codesList[str]

List of API codes - contracts.

APIGenerator

Returns the trained APIGenerator object.

promptmeteo.base module#

class promptmeteo.base.Base(language: str, model_name: str, model_provider_name: str, model_provider_token: str | None = None, model_params: Dict | None = None, prompt_domain: str | None = '', prompt_labels: List[str] = None, prompt_detail: str | None = None, selector_k: int = 10, selector_algorithm: str = 'relevance', verbose: bool = False, **kwargs)#

Bases: ABC

Promptmeteo is a tool powered by LLMs, capable of solving NLP tasks such as text classification and Named Entity Recognition. Its interface resembles that of a conventional ML model, making it easy to integrate into MLOps pipelines.

Notes:#

‘Sun is setting on the New Republic. It’s time for the ResistencIA to rise’

Padmé Amidala, mother of Leia

SELECTOR_TYPE: str = ''#
TASK_TYPE: str = ''#
property builder: TaskBuilder#

Get the TaskBuilder instance for the model.

create_builder() → TaskBuilder#

Create a TaskBuilder instance for the model.

property init_params#

Get the initialization parameters of the model.

property is_trained: bool#

Check if the model is trained.

classmethod load_model(model_path: str) → Self#

Load a saved model from disk.

Parameters:

  • model_path (str) – Path from where the model will be loaded.

  • Returns –

  • ------- –

  • Base – Loaded model instance.

predict(examples: List[str]) → List[str]#

Predict over new text samples.

Parameters:

  • examples (List[str]) – List of text samples to predict.

  • Returns –

  • ------- –

  • List[str] – List of predictions.

save_model(model_path: str) → Self#

Save the trained model to disk.

Parameters:

  • model_path (str) – Path where the model will be saved.

  • Returns –

  • ------- –

  • Base – Instance of the saved model.

property task: Task#

Get the Task instance for the model.

class promptmeteo.base.BaseSupervised(**kwargs)#

Bases: Base

Base class for supervised training tasks.

SELECTOR_TYPE: str = 'supervised'#
train(examples: List[str], annotations: List[str]) → Self#

Trains the model given use cases and notes on behaviour. Check the parameters and task behaviour in each specific model training docstring.

class promptmeteo.base.BaseUnsupervised(**kwargs)#

Bases: Base

Base class for unsupervised training tasks.

SELECTOR_TYPE: str = 'unsupervised'#
train(examples: List[str]) → Self#

Trains the model given use cases and notes on behaviour. Check the parameters and task behaviour in each specific model training docstring.

promptmeteo.code_generator module#

class promptmeteo.code_generator.CodeGenerator(**kwargs)#

Bases: BaseSupervised

Code Generator Task.

This class represents a model for generating code based on natural language descriptions.

Parameters:

  • **kwargs (dict) – Additional keyword arguments.

  • Example –

  • ------- –

  • CodeGenerator (>>> from promptmeteo import) –

  • CodeGenerator( (>>> model =) –

  • language="en" (...) –

:param : :param … prompt_domain=”python”: :param : :param … model_provider_name=”openai”: :param : :param … model_name=”text-davinci-003”: :param : :param … model_provider_token=model_token: :param : :param … prompt_detail=[: :param … “add docstring in function definitions”: :param : :param … “add argument typing annotations”: :param : :param … ]: :param : :param … ): :param >>> pred = model.predict(: :param … [“A function that receives the argument foo and prints it.”]: :param … ): :param Returns: :param ——-: :param None:

TASK_TYPE: str = 'code-generation'#

promptmeteo.constants module#

promptmeteo.document_classifier module#

class promptmeteo.document_classifier.DocumentClassifier(**kwargs)#

Bases: BaseSupervised

DocumentClassifier Task

This class represents a model for classifying documents into predefined categories.

Example:#

>>> from promptmeteo import DocumentClassifier
>>> clf = DocumentClassifier(
...     model_provider_name="hf_pipeline",
...     model_name="google/flan-t5-small",
...     prompt_labels=["positive", "negative", "neutral"],
... )

>>> clf.train(
...     examples=["estoy feliz", "me da igual", "no me gusta"],
...     annotations=["positive", "neutral", "negative"],
... )

>>> clf.predict(["que guay!!"])

>>> [["positive"]]
TASK_TYPE: str = 'classification'#
train(examples: List[str], annotations: List[str]) → Self#
Trains the model given use cases and notes on behaviour. Check the

parameters and task behaviour in each specific model training docstring.

Trains the DocumentClassifier model.

examplesList[str]

List of document examples.

annotationsList[str]

List of corresponding annotations.

Self

promptmeteo.document_qa module#

class promptmeteo.document_qa.DocumentQA(**kwargs)#

Bases: BaseUnsupervised

Question Answering over Documents Task

This class represents a model for answering questions based on documents.

Example:#

>>> from promptmeteo import DocumentQA
>>> clf = DocumentQA(
...     language="en",
...     model_provider_name="hf_pipeline",
...     model_name="google/flan-t5-small",
... )

>>> clf.train(
...     examples=[
...         "The rain in spain is always in plain",
...         "The logarithm's limit is the limit's logarithm",
...         "To punish oppresors is clementy. To forgive them is cruelty",
...     ],
... )

>>> clf.predict(["How is the rain in spain?"])

>>> [["in plain"]]
TASK_TYPE: str = 'qa'#

promptmeteo.exceptions module#

promptmeteo.summarizer module#

class promptmeteo.summarizer.Summarizer(**kwargs)#

Bases: BaseUnsupervised

Class for text summarization

This class represents a model for text summarization.

Example:#

>>> from promptmeteo import Summarizer
>>> model = Summarizer(
...     language="es",
...     prompt_domain="A partir del siguiente texto:",
...     model_name="anthropic.claude-v2",
...     model_provider_name="bedrock",
... )

>>> model.predict([text])
TASK_TYPE: str = 'summarization'#
classmethod load_model(model_path: str) → Self#

Load a saved model from disk.

model_pathstr

Path from where the model will be loaded.

Base

Loaded model instance.

Loads a model artifact to make new predictions.

model_pathstr

The path to the saved model artifact.

selfSummarizer

The loaded Summarizer model.

train() → Self#
Trains the model given use cases and notes on behaviour. Check the

parameters and task behaviour in each specific model training docstring.

Train the Summarizer model.

selfSummarizer

The trained Summarizer model.

promptmeteo.tools module#

promptmeteo.tools.add_docstring_from(parent_function)#

Decorator to concatenate a docstring from another function at the beginning.

Parameters:

  • parent_function (function) – The function whose docstring will be used.

  • Returns –

  • ------- –

  • decorator (function) – The decorator function.

  • Example –

  • ------- –

  • foo() (>>> def) –

  • foo''' (... '''documentation for) –

  • pass (...) –

  • @add_docstring_from(foo) (>>>) –

  • bar() (>>> def) –

  • bar''' (... '''additional notes for) –

  • pass –

  • print(bar.__doc__) (>>>) –

  • foo (>>> documentation for) –

  • bar (>>> additional notes for) –

promptmeteo.validations module#

promptmeteo.validations.validate_version_rest(api_version)#

Validates the version number for REST protocol.

Parameters:

  • api_version (str) – The version number to validate.

  • Returns –

  • ------- –

  • bool – True if the version number is valid, False otherwise.

promptmeteo.validations.version_validation(api_version, api_protocol)#

Validates the version based on the provided protocol.

Parameters:

  • api_version (str) – The version number to validate.

  • api_protocol (str) – The protocol to use for validation.

  • Returns –

  • ------- –

  • bool – True if the version number is valid for the given protocol, False otherwise.

  • Raises –

  • ------ –

  • ValueError – If the provided protocol is not supported.

Module contents#