veris-ai 1.21.0

A Python package for Veris AI tools

Veris AI Python SDK

Tool mocking and MCP server integration for AI agent simulation. For more information visit veris.ai.

Quick Start

Install the SDK:

uv add veris-ai --extra fastapi --extra agents

Minimal Example

from fastapi import FastAPI
from pydantic import BaseModel
from agents import Agent, function_tool
from veris_ai import veris, Runner

app = FastAPI()

# 1. Define tools with @veris.mock() - returns simulated responses during simulation
@function_tool
@veris.mock()
def get_user(user_id: str) -> dict:
    """Get user information."""
    return db.get_user(user_id)

# 2. Create an agent with the tools
agent = Agent(name="Assistant", model="gpt-4", tools=[get_user])

# 3. Expose an endpoint that invokes the agent
class ChatRequest(BaseModel):
    message: str

@app.post("/chat")
async def chat(req: ChatRequest):
    result = await Runner.run(agent, req.message)
    return {"response": result.final_output}

# 4. Set up the MCP server
veris.set_fastapi_mcp(fastapi=app)
veris.fastapi_mcp.mount_http()

Your FastAPI app now exposes an MCP server at /mcp. During Veris simulations, decorated tools return simulated responses; in production, they execute normally.

Installation

# Base package
uv add veris-ai

# With FastAPI MCP support
uv add veris-ai --extra fastapi

# With OpenAI agents support
uv add veris-ai --extra agents

# All extras
uv add veris-ai --extra dev --extra fastapi --extra observability --extra agents

Configuration

Variable	Purpose	Default
VERIS_API_KEY	API authentication key	None
VERIS_MOCK_TIMEOUT	Request timeout (seconds)	90.0

---

MCP Server

The SDK wraps fastapi-mcp to expose your FastAPI endpoints as MCP tools with automatic session handling.

Basic Setup

from fastapi import FastAPI
from veris_ai import veris

app = FastAPI()

# Minimal setup
veris.set_fastapi_mcp(fastapi=app)
veris.fastapi_mcp.mount_http()

Configuration Options

veris.set_fastapi_mcp(
    fastapi=app,

    # Server metadata
    name="My API Server",
    description="API for user management",

    # Filter which endpoints become MCP tools
    include_operations=["get_user", "update_user"],  # Only these operation IDs
    # OR
    exclude_operations=["internal_cleanup"],          # All except these

    # Response schema options
    describe_all_responses=True,        # Include all response schemas
    describe_full_response_schema=True, # Include full JSON schema
)

Filtering Rules:

• Cannot use both include_operations and exclude_operations

Accessing the MCP Server

The MCP server is available at /mcp on your FastAPI base URL:

• Local: http://localhost:8000/mcp
• Production: https://your-api.com/mcp

---

Tool Mocking

Decorators control function behavior during simulations.

@veris.mock()

Returns simulated responses when a session is active:

from veris_ai import veris

@veris.mock()
async def get_account_balance(account_id: str) -> dict:
    """Get account balance."""
    return await bank_api.get_balance(account_id)

Options:

@veris.mock(
    mode="tool",           # "tool" (default) or "function"
    expects_response=True, # Whether to wait for mock response
    cache_response=False,  # Cache responses for identical calls
)

@veris.spy()

Executes the original function and logs the call/response:

@veris.spy()
async def process_payment(amount: float, recipient: str) -> dict:
    """Process a payment - logs but executes normally."""
    return await payment_api.send(amount, recipient)

@veris.stub()

Returns a fixed value during simulations:

@veris.stub(return_value={"status": "success", "id": "test-123"})
async def create_order(items: list) -> dict:
    """Create order - returns stub in simulation."""
    return await order_api.create(items)

Decorator Behavior Summary

Decorator	Session Active	No Session
@veris.mock()	Returns simulated response	Executes normally
@veris.spy()	Executes and logs	Executes normally
@veris.stub()	Returns fixed value	Executes normally

---

Veris Runner

For OpenAI Agents, use the Veris Runner to intercept tool calls without modifying tool code.

Installation

uv add veris-ai --extra agents

Basic Usage

from veris_ai import Runner
from agents import Agent, function_tool

@function_tool
def calculator(x: int, y: int) -> int:
    """Add two numbers."""
    return x + y

agent = Agent(
    name="Assistant",
    model="gpt-4",
    tools=[calculator],
)

# Use Veris Runner instead of OpenAI's Runner
result = await Runner.run(agent, "What's 10 + 5?")

Selective Tool Interception

from veris_ai import Runner, VerisConfig

# Only intercept specific tools
config = VerisConfig(include_tools=["calculator", "search"])
result = await Runner.run(agent, "Calculate 2+2", veris_config=config)

# Or exclude specific tools
config = VerisConfig(exclude_tools=["get_weather"])
result = await Runner.run(agent, "Check weather", veris_config=config)

Per-Tool Configuration

from veris_ai import Runner, VerisConfig, ToolCallOptions, ResponseExpectation

config = VerisConfig(
    tool_options={
        "calculator": ToolCallOptions(
            response_expectation=ResponseExpectation.REQUIRED,
            cache_response=True,
        ),
        "search": ToolCallOptions(
            response_expectation=ResponseExpectation.NONE,
        ),
    }
)

result = await Runner.run(agent, "Calculate and search", veris_config=config)

---

Development

# Install with dev dependencies
uv add veris-ai --extra dev

# Lint and format
ruff check --fix .
ruff format .

# Run tests
pytest tests/ --cov=veris_ai

# Type check
mypy src/veris_ai

---

License

MIT License - see LICENSE for details.