convez.ai
PluginIntegração & APIsIniciante
iam-abbas

Limeprompt

iam-abbas/limeprompt

Prompting leve e estruturado para LLMs com validação Pydantic

Ver no GitHub
InicialPluginScore · de 0 a 100

Biblioteca Python leve para prompting e parsing estruturado com modelos de linguagem. Gera outputs formatados em Pydantic models a partir de prompts parametrizados, com suporte a Claude e OpenAI. Inclui chain-of-thought opcional e logging configurável.

Instalação

$npx convez-ai iam-abbas/limeprompt

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
0
Acompanhando
1
Saúde
28%
Último commit
há 2 anos
Versão
v0.2.3
Issues abertas
1
Contribuidores
1

Casos de uso

  • Gerar emails ou documentos estruturados com LLMs
  • Extrair dados estruturados de respostas de modelos
  • Prototipar workflows de prompting com validação automática

Veredito

Ferramenta minimalista e bem focada para parsing estruturado com LLMs. Boa escolha se você precisa apenas de outputs validados sem complexidade de multi-agent, mas faltam features mais avançadas de prompting.

README

A tradução para o português deste README ainda está na fila.

Limeprompt 🍋

PyPI version

Lightweight prompting and parsing library for LLM models.

What is Limeprompt?

Limeprompt is an opinionated and lightweight prompting and parsing library for LLM models. It aims to make it easy to generate structured outputs from language models. The library is designed to be simple to use, with a single use-case in mind: generating structured outputs from language models. There wont be any support for multi-agent or complex prompting use-cases.

Installation

pip install limeprompt

Example Usage

Here's a simple example using Anthropic's Claude:

import logging
from anthropic import Anthropic
from pydantic import BaseModel
from limeprompt import Limeprompt

# Define your output structure
class Email(BaseModel):
    subject: str
    message: str

# Set up your Anthropic client
anthropic_client = Anthropic(api_key='your-api-key')

# Create a Limeprompt instance
lp = Limeprompt(
    model_client=anthropic_client,
    model_name='claude-3-5-sonnet-20240620',
    prompt="Write an email to <name> about <topic>",
    variables={"name": "Alice", "topic": "limes"},
    output_model=Email,
    max_tokens=1024,
    include_chain_of_thought=True,  # Set to False to disable chain of thought
    log_level=logging.INFO  # Set the desired log level
)

# Run and get your results
result = lp.run()

print(f"Subject: {result.output.subject}")
print(f"Message: {result.output.message}")
if result.chain_of_thought:
    print(f"\nChain of Thought:\n{result.chain_of_thought}")

Here's an example using OpenAI:

import logging
from openai import OpenAI
from pydantic import BaseModel
from limeprompt import Limeprompt

# Define your output structure
class Email(BaseModel):
    subject: str
    message: str

# Set up your OpenAI client
openai_client = OpenAI(api_key='your-api-key')

# Create a Limeprompt instance
lp = Limeprompt(
    model_client=openai_client,
    model_name='gpt-3.5-turbo',
    prompt="Write an email to <name> about <topic>",
    variables={"name": "Bob", "topic": "lemons"},
    output_model=Email,
    max_tokens=1024,
    include_chain_of_thought=False,  # Disable chain of thought
    log_level=logging.WARNING  # Set log level to WARNING
)

# Run and get your results
result = lp.run()

print(f"Subject: {result.output.subject}")
print(f"Message: {result.output.message}")

Contributing

You are welcome to open issues or submit PRs. Here's my todo list for the library:

  • Add support for OpenAI
  • Add logging options
  • Add ability to disable chain of thought
  • Modularize the prompting techniques
  • Add support for few-shot prompting

License

Limeprompt is released under the MIT License. Feel free to use it in your projects.

Ver o original no GitHub

Bom para

Relacionados