assertllm 0.3.0

The pytest for LLMs — fast, Pydantic-based assertions for AI applications

assertllm

The pytest for LLMs. Test your AI outputs like you test your code.

Quick Start

pip install assertllm[anthropic]

from assertllm import expect, llm_test

@llm_test(
    expect.contains("Paris"),
    expect.latency_under(2000),
    expect.cost_under(0.001),
    model="claude-sonnet-4-6",
)
def test_capital(llm):
    llm("What is the capital of France?")

pytest test_capitals.py -v

test_capitals.py::test_capital
  AI response: "The capital of France is Paris."
  ✓ contains("Paris")
  ✓ latency_under(2000) — 823ms
  ✓ cost_under(0.001) — $0.000023
  PASSED

────────── assertllm summary ──────────
  LLM tests: 1 passed
  Assertions: 3/3 passed
  Total cost: $0.000023
  Avg latency: 823ms

Features

• 22+ assertions — text, performance, agent, composable
• Zero LLM calls for most checks — deterministic and instant
• Built on Pydantic — auto-validation, JSON serialization, schema generation
• Multi-provider — OpenAI, Anthropic, Ollama out of the box
• Agent testing — tool calls, loop detection, call ordering
• Retry support — handle non-deterministic outputs
• CI/CD native — pytest markers, JUnit XML, JSON reporters
• Pydantic AI integration — test your existing agents

Installation

pip install assertllm[anthropic]    # Anthropic
pip install assertllm[openai]       # OpenAI
pip install assertllm[ollama]       # Ollama (local)
pip install assertllm[all]          # Everything

Documentation

Full docs at docs.assertllm.dev

Contributing

git clone https://github.com/bahadiraraz/llmtest
cd llmtest
uv sync --all-extras
uv run pytest tests/ -v

Roadmap

This library is under active development. More providers (Gemini, DeepSeek, Mistral, etc.) and additional assertion types are coming soon.

License

MIT