parsec-llm 0.1.2

Lightweight orchestration toolkit to generate, validate, repair and enforce structured output from LLMs

parsec

Lightweight orchestration toolkit to generate, validate, repair and enforce structured output from large language models (LLMs). The project provides a provider-agnostic adapter interface, validators (JSON/pydantic), and an enforcement engine that retries and repairs LLM output until it conforms to a schema.

This repository contains:

• Adapter abstractions and a concrete OpenAI adapter.
• Validation and repair utilities for JSON output.
• An EnforcementEngine that generates, validates, repairs, and retries.
• Examples and tests demonstrating usage.

Features

• Provider-agnostic adapter interface for plugging different LLMs.
• Native-structured output support (when providers allow JSON responses).
• JSON validation with schema-based repair heuristics.
• Retry loop with feedback to the model for progressive repair.
• Small test suite and example runner using the OpenAI adapter.

Installation

pip install parsec-llm

Or for development:

git clone https://github.com/olliekm/parsec.git
cd parsec
pip install -e ".[dev]"

Quick Example

from parsec.models.adapters import OpenAIAdapter
from parsec.validators import JSONValidator
from parsec.enforcement import EnforcementEngine

# Set up components
adapter = OpenAIAdapter(api_key="your-api-key", model="gpt-4o-mini")
validator = JSONValidator()
engine = EnforcementEngine(adapter, validator)

# Define your schema
schema = {
    "type": "object",
    "properties": {
        "name": {"type": "string"},
        "age": {"type": "integer"}
    },
    "required": ["name"]
}

# Enforce structured output
result = await engine.enforce(
    "Extract: John Doe is 30 years old",
    schema
)

print(result.parsed_output)  # {"name": "John Doe", "age": 30}

Development Setup

Requirements: Python 3.9+

1. Install dependencies:

pip install -e ".[dev]"

2. Run tests:

poetry run pytest -q

3. Run the OpenAI example (requires OPENAI_API_KEY):

export OPENAI_API_KEY="sk-..."
export OPENAI_MODEL="gpt-4o-mini"  # optional
poetry run python examples/run_with_openai.py

The example demonstrates using OpenAIAdapter, JSONValidator and EnforcementEngine to extract structured data using a JSON schema.

Code Structure

• src/parsec/core — core abstractions and schemas.
• src/parsec/models — provider adapters (OpenAI, Anthropic).
• src/parsec/validators — validator implementations.
• src/parsec/enforcement — enforcement/orchestration logic.
• src/parsec/cache — caching implementations.
• src/parsec/utils — utility functions (partial JSON parsing).
• examples/ — example runners (real OpenAI example included).
• tests/ — unit tests with pytest.

Testing

Run the test suite with:

poetry run pytest -q

Notes & Next Steps

• The OpenAI example performs real API calls — be mindful of API keys and costs when running it.
• Consider mocking adapters for offline or CI-safe tests.
• This project is intentionally minimal and modular — adapters and validators can be extended to support additional providers and formats.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Copyright (c) 2025 Oliver Kwun-Morfitt