LLMs Are Fluent, Confident, and Totally Wrong

Until Now.

llm-contracts is the first developer-first framework for validating, linting, and asserting the correctness of LLM-generated outputs. Think of it as ESLint + Pytest for AI responses — without requiring a specific model or cloud API.

Get Started on GitHub Read the Whitepaper

The Problem That Keeps You Up At Night

You deployed an AI feature last week. It’s working perfectly.
Except when it confidently tells customers about warranty policies that don’t exist.
Or invents product specifications.
Or promises shipping to countries you don’t serve.

Sound familiar?

Real-World AI Failures:

Air Canada’s chatbot promised non-existent bereavement fares → Legal action
CNET’s AI writer published financial advice with wrong interest rates → Public corrections
ChatGPT lawyer submitted fake legal citations in court → Professional sanctions
E-commerce AI invented product features that didn’t exist → Customer complaints

The pattern: LLMs generate responses that look perfect but break everything.

The Solution: Contracts for AI Output

Instead of hoping your AI “gets it right,” llm-contracts lets you define exactly what “right” looks like:

# Your contract with the AI
schema:
  warranty_period:
    type: str
    pattern: "^(30|90|365) days?$"  # Only valid warranty periods
  
rules:
  - keyword_must_include: ["warranty", "return policy"]
  - keyword_must_not_include: ["guaranteed", "always", "never"]
  - no_placeholder_text: "\\[INSERT_.*\\]"
  - word_count_min: 100
  - phrase_proximity:
      terms: ["warranty", "30 days"]
      max_distance: 20  # Warranty details must be close together

Result: Every AI response gets validated before reaching your users. No more silent failures.

See It In Action

Before llm-contracts:

{
  "product_description": "[INSERT_PRODUCT_NAME] is the best quality item you'll ever buy! We guarantee 100% satisfaction always and forever. Our unlimited warranty covers everything!"
}

Passes the “looks good” test. Breaks everything else.

After llm-contracts:

Schema validation: PASSED
Placeholder text detected: "[INSERT_PRODUCT_NAME]"
Forbidden keywords: "guarantee", "always", "unlimited"
Missing required: "30-day warranty", "return policy"

VALIDATION FAILED - Output rejected

Catches the problems before they reach production.

Why Developers Choose llm-contracts

Framework-Agnostic

Works with any LLM (OpenAI, Anthropic, local models) and any framework (LangChain, direct API calls, custom implementations).

Production-Ready

Built by developers who’ve shipped AI features at scale. Handles edge cases, provides detailed error reporting, and integrates with your existing CI/CD.

Zero Vendor Lock-in

No API calls to external services. No model-specific prompting tricks. Just pure validation logic that runs anywhere Python runs.

Professional Reports

Generate beautiful HTML and Markdown validation reports for stakeholders, compliance, and debugging.

How Is This Different?

Validate AI like you validate code — enforce rules, not hope for the best.

Tool	Approach	llm-contracts Difference
Guardrails	Tries to fix bad output	We fail fast — no magic repairs
LangChain	Orchestrates AI pipelines	We’re a QA layer, not orchestration
Pydantic	Python type validation	We validate content quality, not just types
Manual Review	Human checks everything	We automate validation at machine speed

Other tools try to fix LLM output.
llm-contracts asks: “Did the AI follow the rules?”
If not, we fail it — no excuses.

Get Started Today

Installation

pip install llm-contracts

Quick Validation

# Validate AI output against your rules
llm-validate output.json --schema schema.yaml --html-report report.html

Python API

from llm_contracts import contracts

# Validate and get detailed results
result = contracts.validate(ai_output, "schema.yaml")
if not result.is_valid:
    print(f"AI failed validation: {result.errors}")

View Full Documentation Try the Web Interface

Learn More

The Problem - Why LLM validation matters
Core Features - Schema validation, content linting, and reporting
Real Use Cases - E-commerce, support, marketing, and more
Web Interface - Try validation in your browser
Roadmap - Upcoming features and development plans
Complete Whitepaper - Deep dive into the business case

Stop hoping your AI gets it right. Start knowing it does.

Get Started on GitHub →

Created by Mohamed Jama for developers who ship AI features that actually work.

Major contributions by Abdirahman Attila - Frontend interface, documentation website, and testing infrastructure.