DFF Universal Prompted Skill

Description

This skill is aimed for debugging purposes, it utilizes the given in the request to the dialog system (!) LLM service to generate a response to the given dialog context and prompt.

Use it in the following way:

1. raise a dialog system with the skill
2. send a request to it from python in the folllowing way:

import random
import requests

prompt = """Your prompt here"""
lm_service = "http://openai-api-chatgpt:8145/respond"  # or any other from the added to your distribution
UNIVERSAL_ASSISTANT = "http://0.0.0.0:4242"  # if the agent raised locally on the same server

result = requests.post(
     UNIVERSAL_ASSISTANT, 
     json={
         "user_id": f"test-user-000", 
         "payload": "Who are you? who built you? what can you do?",
         "prompt": prompt, 
         "lm_service_url": lm_service,
         "openai_api_key": "Your OpenAI API Key here",
     }).json()

Parameters

GENERATIVE_TIMEOUT: timeout for request to LLM
N_UTTERANCES_CONTEXT: number of last utterances to consider as a dialog context
DEFAULT_LM_SERVICE_URL: default LLM to utilize if not provided in the request
DEFAULT_LM_SERVICE_CONFIG: onfiguration file with generative parameters to utilize if not provided in the request

Dependencies

• LLM GENERATIVE_SERVICE_URL
• API keys in environmental variables for key-required LLMs (OpenAI API, Anthropic API)