​
Integrating Prompt Templates into Spans

By instrumenting the prompt template, users can fully utilize Future AGI’s prompt playground. There’s no need to deploy a new template version to test if changes in prompt text or variables achieve the desired effect. Instead, you can experiment with these modifications directly in the playground UI.

​
Implementation Details

We provide a using_prompt_template context manager to add a prompt template into the current OpenTelemetry Context. FI auto-instrumentors will read this Context and pass the prompt template fields as span attributes, adhering to the traceAI semantic conventions.

​
Required Parameters

ParameterTypeDescriptionExample
templatestrThe string for the prompt template”Please describe the weather forecast for on “
versionstrIdentifier for the template version”v1.0”
variablesDict[str]Dictionary containing variables to fill the template{"city": "San Francisco", "date": "March 27"}

​
Sample Implementation

Begin by installing the necessary dependencies:

pip install fi-instrumentation-otel traceai_openai openai

Below is a comprehensive example demonstrating how to implement prompt template tracing:

import os
import openai
import opentelemetry
from fi_instrumentation import register, using_prompt_template
from openai import OpenAI
from traceai_openai import OpenAIInstrumentor

# Set up Environment Variables
os.environ["OPENAI_API_KEY"] = "your-openai-api-key"
os.environ["FI_API_KEY"] = "your-futureagi-api-key"
os.environ["FI_SECRET_KEY"] = "your-futureagi-secret-key"

my_first_model = "my first model"

# Setup OTel via our register function
trace_provider = register(
    project_type=ProjectType.EXPERIMENT,
    project_name="Project_name",
    project_version_name="project_version_name",
)
OpenAIInstrumentor().instrument(tracer_provider=trace_provider)

# Setup OpenAI
client = OpenAI()

# Define the prompt template and its variables
prompt_template = "Please describe the weather forecast for {city} on {date}"
prompt_template_variables = {"city": "San Francisco", "date":"March 27"}

# Use the context manager to add template information
with using_prompt_template(
    template=prompt_template,
    variables=prompt_template_variables,
    version="v1.0",
):
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {
                "role": "user",
                "content": prompt_template.format(**prompt_template_variables)
            },
        ]
    )