Promptify is a task-based NLP engine that uses large language models to generate structured outputs without requiring training data.
The tool solves the problem of extracting structured information from text using LLMs by wrapping common NLP tasks behind simple interfaces that return Pydantic-validated objects instead of raw strings. It works by accepting text input, sending it to an LLM provider through LiteLLM's unified backend, and parsing the response into a predefined schema. Built-in tasks include named entity recognition, classification, question answering, summarization, relation extraction, and SQL generation. For problems outside these categories, developers can define custom tasks using any Pydantic schema. The tool supports few-shot examples to improve accuracy, domain specialization through context hints, and includes an evaluation framework with metrics like precision, recall, F1, and ROUGE scores.
Promptify suits developers who need structured outputs from LLMs without building custom prompt engineering infrastructure. It works well for projects that can tolerate LLM latency and cost, need multiple NLP capabilities, or want to experiment with different LLM providers without rewriting code. The tool's main advantage is reducing boilerplate: common tasks require only a few lines of code. It supports any LLM provider via LiteLLM, including OpenAI, Anthropic, Google, Ollama, and Azure. The project includes batch processing with async concurrency, native async/await support, cost tracking via token usage monitoring, and a safe JSON parser that falls back to completion-based parsing for providers without native structured output support.
Development activity shows consistent engagement with the community through a Discord channel for prompt engineering discussions. The project maintains an open contribution policy and actively incorporates feedback from users.
