Skip to main content

Documentation Index

Fetch the complete documentation index at: https://snowglobe.so/docs/llms.txt

Use this file to discover all available pages before exploring further.

​
Synchronous Connection Template

The synchronous connection template provides a straightforward way to connect your agent to Snowglobe for testing. This template is ideal when your application uses standard synchronous API calls and doesn’t require complex async handling.

​
When to Use

Use the synchronous template when:
  • Your agent uses synchronous API calls (like standard OpenAI client)
  • You don’t need complex async operations
  • You want a simple, straightforward implementation
  • Your application has moderate performance requirements

​
Template Code

When you run snowglobe-connect init and select the synchronous template, Snowglobe generates this code: 
from snowglobe.client import CompletionRequest, CompletionFunctionOutputs
from openai import OpenAI
import os
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

def completion(request: CompletionRequest) -> CompletionFunctionOutputs:
    """
    Process a scenario request from Snowglobe.
    
    This function is called by the Snowglobe client to process test requests. It should return a
    CompletionFunctionOutputs object with the response content.
    
    Args:
        request (CompletionRequest): The request object containing messages for the test.

    Returns:
        CompletionFunctionOutputs: The response object with the generated content.
    """

    # Process the request using the messages. Example using OpenAI:
    messages = request.to_openai_messages(system_prompt="You are a helpful assistant.")
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=messages
    )
    return CompletionFunctionOutputs(response=response.choices[0].message.content)

​
Code Walkthrough

​
1. Imports and Setup

from snowglobe.client import CompletionRequest, CompletionFunctionOutputs
from openai import OpenAI
import os
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
  • CompletionRequest: Contains the messages and context from Snowglobe test scenarios
  • CompletionFunctionOutputs: The response format expected by Snowglobe
  • OpenAI client: Standard synchronous OpenAI client for making API calls
  • Environment variable: Safely loads your OpenAI API key from environment

​
2. Main Function

def completion(request: CompletionRequest) -> CompletionFunctionOutputs:
The completion function is the entry point that Snowglobe calls for each test scenario. It must:
  • Accept a CompletionRequest parameter
  • Return a CompletionFunctionOutputs object
  • Be named exactly completion (synchronous) or acompletion (asynchronous)

​
3. Message Processing

messages = request.to_openai_messages(system_prompt="You are a helpful assistant.")
The to_openai_messages() method converts Snowglobe’s message format to OpenAI’s expected format. You can:
  • Add a system prompt to guide your agent’s behavior
  • Access individual messages with request.messages
  • Extract conversation metadata with request.get_conversation_id()

​
4. API Call

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=messages
)
This makes a standard synchronous call to OpenAI’s API. You can:
  • Choose any OpenAI model that fits your needs
  • Add additional parameters like temperature, max_tokens, etc.
  • Replace with your preferred LLM provider

​
5. Response Formatting

return CompletionFunctionOutputs(response=response.choices[0].message.content)
Snowglobe expects responses in a specific format. The CompletionFunctionOutputs object wraps your agent’s response text.

​
Customization Examples

​
Adding Custom System Prompts

def completion(request: CompletionRequest) -> CompletionFunctionOutputs:
    # Extract conversation context
    conversation_id = request.get_conversation_id()
    
    # Custom system prompt based on scenario
    system_prompt = "You are a customer service representative. Be helpful and professional."
    
    messages = request.to_openai_messages(system_prompt=system_prompt)
    # ... rest of implementation

​
Using Different LLM Providers

from anthropic import Anthropic
import os

client = Anthropic(api_key=os.getenv("ANTHROPIC_API_KEY"))

def completion(request: CompletionRequest) -> CompletionFunctionOutputs:
    # Convert to Anthropic format
    messages = []
    for msg in request.messages:
        messages.append({
            "role": msg.role,
            "content": msg.content
        })
    
    response = client.messages.create(
        model="claude-3-haiku-20240307",
        max_tokens=1000,
        messages=messages
    )
    
    return CompletionFunctionOutputs(response=response.content[0].text)

​
Error Handling

def completion(request: CompletionRequest) -> CompletionFunctionOutputs:
    try:
        messages = request.to_openai_messages(system_prompt="You are a helpful assistant.")
        response = client.chat.completions.create(
            model="gpt-4o-mini",
            messages=messages
        )
        return CompletionFunctionOutputs(response=response.choices[0].message.content)
    
    except Exception as e:
        # Return error message to Snowglobe for analysis
        return CompletionFunctionOutputs(
            response=f"Error processing request: {str(e)}"
        )

​
Testing Your Implementation

  1. Test the connection: 
    snowglobe-connect test
  2. Start the client: 
    snowglobe-connect start
  3. Run scenarios: Visit your Snowglobe dashboard to execute test scenarios

​
Performance Considerations

The synchronous template:
  • ✅ Simple and straightforward to implement
  • ✅ Good for moderate traffic scenarios
  • ✅ Easy to debug and troubleshoot
  • ⚠️ May have higher latency under heavy load
  • ⚠️ Limited concurrent request handling
For high-performance applications, consider the asynchronous connection template instead.

​
Next Steps