Fetch prompts

This guide shows you how to fetch the deployed version of your prompt in your code. You can do this using the Agenta SDK (Python) or the REST API.

Fetching prompts with the Agenta SDK

Step 1: Setup

Make sure to install the latest version of the agenta Python SDK (pip -U install agenta).

• Set up environment variables:
  • AGENTA_API_KEY for cloud users.
  • AGENTA_HOST set to http://localhost if you are self-hosting.
  • AGENTA_PROJECT_ID set to the project ID.

Step 2: Fetch the prompt

import agenta as ag
from agenta import DeploymentManager, PromptManager
ag.init()

# fetch the deployment (id)
deployment = DeploymentManager.get_deployment_by_name(deployment_name="production",
                                                      app_slug="my-app")
# fetch latest configuration deployed here
prompt_obj = PromptManager.get_from_registry(deployment_id=deployment.id)

The prompt object is a dictionary that will have both the information about the configuration version and the data (under the field data)

{
 variant_slug: "my-variant",
 app_slug: "my-app""
 version: 3,
 commit-id: "afae3232",
 data: {
    'temperature': 1.0,
    'model': 'gpt-3.5-turbo',
    'max_tokens': -1,
    'prompt_system': 'You are an expert in geography.',
    'prompt_user': 'What is the capital of {country}?',
    'top_p': 1.0,
    'frequence_penalty': 0.0,
    'presence_penalty': 0.0,
    'force_json': 0
}

Fetching prompts with the REST API

@jp-agenta how would that look like?