retrieving OpenAI tokens for monitoring

[jane-warren-cap]

I'm applying guardrails to a chatbot based on gpt-4o mini. I am testing with a pretty basic setup right now, with only input and output rails. The issue I'm facing is that even when verbose logs are enabled (in CLI chat, a server, or just running a single query through a python file) I see that total tokens, prompt tokens, and completion tokens are all 0. I would like to be able to see how many tokens each response and guardrail is taking up--- can anyone help with enabling this?

for context, my config is very simple:
`models:

• type: main
  engine: openai
  model: gpt-4o-mini

instructions:

rails:
input:
flows:
- self check input
output:
flows:
- self check output
- self check hallucination
When running a simple testing script like the one below, the tokens all output as 0. I wonder if they are not being read in correctly?import asyncio
from nemoguardrails import LLMRails, RailsConfig
from nemoguardrails.logging.explain import ExplainInfo
from nemoguardrails.logging.stats import LLMStats
from dotenv import load_dotenv
import os
import json

load_dotenv("")
OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")

config=RailsConfig.from_path("./config")
rails = LLMRails(config, verbose=False)

async def stream_response(messages):
"""takes messages and chunks them out"""
async for chunk in rails.stream_async(messages=messages):
print(chunk, end="")
print()

messages=[{
"role": "user",
"content": "hello! what can you help me with?"
}]

asyncio.run(stream_response(messages))

info = rails.explain()
print(info.colang_history)
print(info.print_llm_calls_summary)`

[jane-warren-cap]

just noticed you resolved exactly this issue in your most recent merge. just pulled updated files from develop branch and set streaming_usage: True in my config, and all is well!!