semantix-ai

Validate what your LLM outputs mean, not just their shape.

---

semantix is a semantic type system for AI outputs. It uses local NLI (Natural Language Inference) to check whether text means what you require -- no API keys, no network calls, ~15ms per check.

pip install semantix-ai

from semantix.testing import assert_semantic

def test_chatbot_is_polite():
    response = my_chatbot("handle angry customer")
    assert_semantic(response, "polite and professional")

On failure you get a clear diagnostic:

AssertionError: Semantic check failed (score=0.12)
  Intent:  polite and professional
  Output:  "You're an idiot for asking that."
  Reason:  Text contains aggressive language

Key properties

• Local inference -- NLI model runs on CPU, no data leaves your machine
• ~15ms per check -- negligible overhead on any LLM call
• Zero API cost -- no tokens burned for validation
• Pluggable judges -- NLI, embedding, LLM, quantized ONNX, forensic, caching
• Framework integrations -- Guardrails AI, Instructor, Pydantic AI, LangChain, DSPy
• Self-healing retries -- structured feedback so the LLM corrects itself
• Training flywheel -- every retry produces labeled fine-tuning data

How it works

1. You define an Intent -- a class whose docstring describes what the text should mean.
2. A Judge evaluates whether the output entails the intent using NLI, embeddings, or an LLM.
3. If validation fails, semantix can retry with structured feedback, and collect the correction pair for fine-tuning.

from semantix import Intent, validate_intent

class ProfessionalDecline(Intent):
    """The text must politely decline an invitation without being rude."""

@validate_intent
def decline_invite(event: str) -> ProfessionalDecline:
    return call_my_llm(event)

result = decline_invite("the company retreat")
# Returns a validated ProfessionalDecline -- or raises SemanticIntentError

Quick links

Topic	Description
Getting Started	Install and run your first semantic test in under 2 minutes
Intents	Define semantic contracts with Intent classes
Composition	Combine intents with Not, AllOf, AnyOf, ~, &, |
assert_semantic	One-line semantic assertions for any test runner
pytest-semantix	Pytest plugin with markers, reports, and CI integration
Judges	NLI, Embedding, LLM, Quantized, Forensic, Caching
Integrations	Guardrails AI, Instructor, Pydantic AI, LangChain, DSPy
Training	Collect correction pairs, calibrate thresholds, export for fine-tuning
Advanced	Streaming, audit trails, MCP server, async

Installation options

pip install semantix-ai                    # Core (default NLI judge)
pip install "semantix-ai[turbo]"           # Quantized ONNX (smallest footprint)
pip install "semantix-ai[openai]"          # LLM judge (GPT-4o-mini)
pip install "semantix-ai[instructor]"      # Instructor integration
pip install "semantix-ai[pydantic-ai]"     # Pydantic AI integration
pip install "semantix-ai[langchain]"       # LangChain integration
pip install "semantix-ai[guardrails]"      # Guardrails AI integration
pip install "semantix-ai[dspy]"            # DSPy integration
pip install "semantix-ai[mcp,nli]"         # MCP server
pip install "semantix-ai[all]"             # Everything

Note

The package name on PyPI is semantix-ai. The import is from semantix import ....