QWED Neurosymbolic Verification
Version updated for https://github.com/QWED-AI/qwed-verification to version v2.3.0.
- This action is used across all versions by 0 repositories.
Go to the GitHub Marketplace to find the latest changes.
Action Summary
QWED Protocol is a model-agnostic verification tool designed to validate outputs from any large language model (LLM) by detecting and preventing hallucinations. It automates the verification of AI-generated content using eight specialized engines, ensuring only provable outputs are used in production. By integrating multiple existing verification libraries into a unified API, QWED provides a reliable solution for enhancing AI trustworthiness in production environments.
Release notes
🦜 Native LangChain Integration - Platform Partner Model
QWED v2.3.0 brings first-party LangChain support as an independent platform partner - following the same model as OpenAI, Anthropic, and Google.
🎯 What’s New
Native LangChain Integration
- ✅
qwed_sdk/integrations/langchainmodule - ✅ QWEDTool with BaseTool interface
- ✅ Optional dependency:
pip install 'qwed[langchain]' - ✅ Full ownership & control - no gatekeepers!
📦 Installation
pip install --upgrade 'qwed[langchain]
🚀 Quick Start
from qwed_sdk.integrations.langchain import QWEDTool
from langchain.agents import initialize_agent
from langchain_openai import ChatOpenAI
tool = QWEDTool(provider="openai", model="gpt-4o-mini")
llm = ChatOpenAI()
agent = initialize_agent(tools=[tool], llm=llm)
agent.run("Verify: what is the derivative of x^2?")
# ✅ VERIFIED: 2*x (100% confidence)
🔗 Combine with v2.2.0 Features
# Enterprise privacy + LangChain
tool = QWEDTool(provider="openai", mask_pii=True)
📊 What’s Included
- Native LangChain BaseTool integration
- Graceful dependency handling
- Pydantic input schemas
- Async support
- Comprehensive error messages
PyPI: https://pypi.org/project/qwed/2.3.0/ Docs: https://github.com/QWED-AI/qwed-verification#-llm-framework-integrations Full Changelog: https://github.com/QWED-AI/qwed-verification/compare/v2.2.0...v2.3.0