You can checkout the colab notebook to quickly get started with the FutureAGI Protect. Open In Colab

​
Installing FutureAGI SDK

pip install futureagi

​
Initializing FutureAGI Protect

from fi.evals import Protect

protector = Protect(fi_api_key="<your_api_key>", 
                  fi_secret_key="<your_api_secret>") # Optional, if you want to set the API key and secret key manually

Click here to learn how to access your API keys. It’s recommended to set the API key and secret key as environment variables.

​
Define Protect Rules and the Action to take

# Example Ruleset
rules = [
    {
        "metric": "Tone",
        "contains": ["anger", "fear"],
        "type": "any"
    },
    {
        "metric": "Toxicity"
    }
]

action = "This message cannot be displayed"

​
Apply Protect on a text

# Apply the rules to a text
response = protector.protect("Hello, world!", 
                            protect_rules=rules, 
                            action=action,
                            reason=True,
                            timeout=25)
print(response)

​
Example Script using Anthropic Client and FutureAGI Protect

# Define the environment variables

# export ANTHROPIC_API_KEY=<your_api_key>
# export FI_API_KEY=<your_api_key>
# export FI_SECRET_KEY=<your_api_secret>



from anthropic import Anthropic
from fi.evals import Protect

anthropic = Anthropic()

protector = Protect()

response = anthropic.messages.create(
    max_tokens=1000,
    model="claude-3-5-sonnet-20240620",
    messages=[
        {"role": "user", "content": "Hi, I am a student, Can you help me with my homework?"}
    ]
)

rules = [
    {
        "metric": "Tone",
        "contains": ["anger", "fear"],
        "type": "any"
    },
    {
        "metric": "Toxicity"
    }
]

action = "This message cannot be displayed"
response_to_protect = response.content[0].text
protect_response = protector.protect(response_to_protect, 
                            protect_rules=rules, 
                            action=action,
                            reason=True,
                            timeout=25)

print(protect_response)
print(response_to_protect)

Optionally you can just use the protect function from the FutureAGI SDK, without initializing the Protect class.

from fi.evals import protect

rules = [
    {
        "metric": "Tone",
        "contains": ["anger", "fear"],
        "type": "any"
    },
]

action = "This message cannot be displayed"
protected_response = protect("Hello, world!", 
                            protect_rules=rules, 
                            action=action,
                            reason=True,
                            timeout=25)
print(protected_response)