Python Runtime

The Prompty Python runtime provides a powerful and flexible way to execute prompts programmatically. It’s designed to be extensible, observable, and easy to integrate into your AI applications.

Installation

Install the Prompty runtime using pip. Choose the appropriate extras based on your needs:

# Basic installation
pip install prompty

# With Azure OpenAI support
pip install "prompty[azure]"

# With OpenAI support
pip install "prompty[openai]"

# With all invokers
pip install "prompty[azure,openai,serverless]"

Basic Usage

Executing Prompty Files

The simplest way to use the runtime is with the execute() function:

import prompty
import prompty.azure  # Import the invoker you need

# Execute a prompty file
response = prompty.execute("path/to/your/prompt.prompty")
print(response)

Passing Input Variables

You can pass variables to your prompts using the inputs parameter:

response = prompty.execute(
    "path/to/your/prompt.prompty",
    inputs={
        "customer_name": "John Doe",
        "question": "What are your business hours?"
    }
)

Core Functions

`execute()`

The main function for running prompts. It combines loading, preparing, and executing in one call.

def execute(
    prompt: Union[str, Prompty],
    *,
    inputs: dict[str, Any] = {},
    connection: str = "default",
    configuration: dict[str, Any] = {},
    options: dict[str, Any] = {},
    stream: bool = False
) -> Any

Parameters:

prompt: Path to prompty file or Prompty object
inputs: Variables to pass to the prompt template
connection: Configuration connection name
configuration: Model configuration overrides
options: Additional options for execution
stream: Whether to stream the response

`load()`

Load a prompty file into a Prompty object:

prompt = prompty.load("path/to/prompt.prompty")
print(f"Loaded prompt: {prompt.name}")

`prepare()`

Prepare inputs and render the prompt template:

prepared = prompty.prepare(
    prompt,
    inputs={"name": "Alice", "topic": "AI"}
)

`run()`

Execute a prepared prompt against the model:

result = prompty.run(prepared)

Headless Usage

For programmatic prompt creation without files:

# Create a headless prompt
prompt = prompty.headless(
    api="chat",
    content="Hello, {{name}}! Tell me about {{topic}}.",
    connection={
        "type": "azure_openai",
        "azure_endpoint": "https://your-endpoint.openai.azure.com/",
        "azure_deployment": "gpt-35-turbo"
    }
)

# Execute it
response = prompty.execute(prompt, inputs={"name": "Bob", "topic": "Python"})

Available Invokers

The runtime supports multiple AI service providers:

Azure OpenAI

import prompty.azure

# Configuration in prompty file or programmatically
response = prompty.execute(
    prompt,
    configuration={
        "type": "azure_openai",
        "azure_endpoint": "${env:AZURE_OPENAI_ENDPOINT}",
        "azure_deployment": "gpt-35-turbo",
        "api_version": "2024-10-21"
    }
)

OpenAI

import prompty.openai

response = prompty.execute(
    prompt,
    configuration={
        "type": "openai",
        "api_key": "${env:OPENAI_API_KEY}",
        "model": "gpt-3.5-turbo"
    }
)

Serverless Models

Support for GitHub Models and other serverless endpoints:

import prompty.serverless

response = prompty.execute(
    prompt,
    configuration={
        "type": "serverless",
        "endpoint": "https://models.inference.ai.azure.com",
        "model": "gpt-4o-mini"
    }
)

Streaming Responses

Enable streaming for real-time response processing:

# Get streaming response
stream = prompty.execute("prompt.prompty", stream=True)

# Process chunks as they arrive
for chunk in stream:
    print(chunk, end="")

Async Support

The runtime provides async versions of all main functions:

import asyncio
import prompty

async def main():
    # Async execution
    response = await prompty.execute_async("prompt.prompty")

    # Async loading
    prompt = await prompty.load_async("prompt.prompty")

    # Async streaming
    async for chunk in await prompty.execute_async("prompt.prompty", stream=True):
        print(chunk, end="")

asyncio.run(main())

Error Handling

Handle common runtime errors:

try:
    response = prompty.execute("prompt.prompty")
except FileNotFoundError:
    print("Prompty file not found")
except ValueError as e:
    print(f"Invalid configuration: {e}")
except Exception as e:
    print(f"Execution failed: {e}")

Best Practices

Next Steps

Learn about Observability & Tracing to monitor your prompts
Explore the CLI Usage for command-line operations
Check out Advanced Configuration for complex scenarios

Want to Contribute To the Project? - Updated Guidance Coming Soon.