Glossary¶

Key terms used throughout the Tenro documentation.

Simulation¶

Replacing real external calls with controlled responses during tests. Tenro simulates LLM provider responses and tool results so you can test without network calls.

Terminology

We use "simulation" rather than "mock" or "mocking" throughout Tenro documentation. Exception: external library names like respx_mock are unchanged.

Construct¶

The test harness providing simulate_* and verify_* methods. Injected as a pytest fixture.

from tenro import Provider
from tenro.simulate import llm, tool
from tenro.testing import tenro

@tenro
def test_agent():  # Construct injected here
    llm.simulate(Provider.OPENAI, response="Hello!")
    llm.verify()

Linking¶

Decorators (@link_tool, @link_llm, @link_agent) that connect your application code to Tenro's simulation and tracing system.

from tenro import link_agent, link_tool

@link_tool
def search(query: str) -> list[str]: ...

@link_agent
def my_agent(query: str) -> str: ...

Evals¶

Evaluation functions for asserting agent behaviour (output quality, tool usage, etc.).

Support levels¶

Supported¶

Tested before every release. We fix breakages promptly.

Examples: OpenAI provider, Anthropic provider, @link_tool decorator.

Experimental¶

Tested, but may break when frameworks update.

Examples: LangChain recipes, CrewAI recipes, framework integration examples.

User personas¶

Framework user¶

Developer using an agent framework (LangChain, CrewAI, LangGraph, Pydantic AI, etc.) where the framework handles LLM calls internally.

Key insight: Framework users cannot use @link_llm because they don't control the LLM call code. Tenro's HTTP interception handles LLM simulation automatically.

Custom agent builder¶

Developer making raw LLM SDK calls (OpenAI, Anthropic, etc.). Controls all code and can use all decorators including @link_llm.

HTTP interception¶

The technique Tenro uses to simulate LLM responses. Tenro intercepts outbound HTTP requests to provider endpoints and returns configured simulated responses.

Compatibility: Works with httpx-based HTTP transports. See Compatibility.