v0.1.0 — Genesis Release

ground-truth v0.1.0 — Genesis Release

Hallucination detection for RAG pipelines. Verify that LLM outputs are grounded in source documents.

What's Included

• Claim extraction from LLM outputs
• Source document matching and verification
• Confidence scoring per claim
• Support for multiple evidence sources
• Integration with popular RAG frameworks
• Batch processing for pipeline use
• 18 commits, comprehensive test coverage

Install

pip install ground-truth

Quick Start

from ground_truth import Verifier

verifier = Verifier()
result = verifier.check(
    claim="The company was founded in 2025",
    sources=["company_docs.pdf", "filing.txt"]
)
print(f"Grounded: {result.is_grounded} ({result.confidence:.2f})")

Built with the Edgecraft Protocol. TechKnowMad Labs — Autonomous Intelligence.