bug: content blocked for any user query by content safety check #1299
Description
Did you check docs and existing issues?
- I have read all the NeMo-Guardrails docs
- I have updated the package to the latest version before submitting this issue
- (optional) I have used the develop branch
- I have searched the existing issues of NeMo-Guardrails
Python version (python --version)
Python 3.11
Operating system/version
Windows
NeMo-Guardrails version (if you must use a specific version and not the latest
No response
Describe the bug
Hi, I am getting content blocked message even for greetings like "Hi!" I used the config/config.yml from
https://docs.nvidia.com/nemo/guardrails/latest/getting-started.html# and the config/prompts.yml using https://docs.nvidia.com/nemo/guardrails/latest/_downloads/581b1286b50c732ffcf3928197e66b67/prompts.yml
My test script looks like the following:
import asyncio
import os
os.environ["NVIDIA_API_KEY"] = # The Nvidia API Key here
from nemoguardrails import LLMRails, RailsConfig
config = RailsConfig.from_path("./config")
rails = LLMRails(config)
async def stream_response(messages):
async for chunk in rails.stream_async(messages=messages):
print(chunk, end="")
print()
messages=[{
"role": "user",
"content": "Hi"
}]
asyncio.run(stream_response(messages))
The response I am getting is:
It's lovely to meet you. I hope you're having a fantastic day so far. My name is Assistron, by the way - I'm an AI designed to assist and communicate with users in a helpful and friendly manner. I can provide information on a wide range of topics, from science and history to entertainment and culture. If you're feeling stuck or need some inspirat{"error": {"message": "Blocked by content safety check output $model=content_safety rails.", "type": "guardrails_violation", "param": "content safety check output $model=content_safety", "code": "content_blocked"}}
What am I doing wrong here?
Steps To Reproduce
Use the config and prompts files from the links provided and run the script above
Expected Behavior
The content is not blocked for greetings and queries which are not harmful
Actual Behavior
Every single query is blocked whether its good or bad