Unsure is a library for creating operations on uncertain or ambiguous values, utilizing an inference endpoint to determine transformations and comparisons. It's meant to leverage ai while yielding predictable and invariant results.
To install the package, run:
pip install py-unsureAll that needs to be configured is the inference endpoint.
It supports Groq apis and OpenAi apis for now, so a groqApiKey can be provided like this
from unsure_ai import config_global_unsure
config_global_unsure(groq_api_key='your key here')and for openAi
import { Unsure, configGlobalUnsure } from 'unsure-js';
configGlobalUnsure({ openAiApiKey: 'your key here' });from unsure_ai import config_global_unsure
def inference_endpoint(q: str) -> str:
# any function that returns a string here, you can call OpenAI, Gemini, Claude, or your own model, just return a string
pass
config_global_unsure(inference_endpoint=inference_endpoint)Once it's configured you can start using the operators just like this
Checks equality. Example:
from unsure_ai import Unsure, config_global_unsure
config_global_unsure(groq_api_key='your key here')
print(Unsure("Lion").is_("Mammal")) # TruePicks an information from a string. Example:
from unsure_ai import Unsure, config_global_unsure
config_global_unsure(groq_api_key='your key here')
print(Unsure("Contact Number: +1-800-555-5555").pick("phone number")) # "+1-800-555-5555"
print(Unsure("Phone: +1-800-555-5555").pick("phone number")) # "+1-800-555-5555"
print(Unsure("Call us at +1-800-555-5555").pick("phone number")) # "+1-800-555-5555"Categorizes the string into the given categories. Example:
from unsure_ai import Unsure, config_global_unsure
config_global_unsure(groq_api_key='your key here')
print(Unsure("Sky").categorize(["blue", "green"])) # "blue"
print(Unsure("Grass").categorize(["blue", "green"])) # "green"Transforms the string into what's demanded. Example:
from unsure_ai import Unsure, config_global_unsure
config_global_unsure(groq_api_key='your key here')
print(Unsure("Response: {\"key\": \"should get this\"}").flat_map_to("key's value")) # "should get this"
print(
Unsure("HTML Content: <html><body><div class=\"scrapable\">Target Content<div></body></html>")
.flat_map_to("content of the div with the class scrapable")
) # "target content"
print(Unsure("Favorite Color: #FF5733").flat_map_to("color in hex")) # "#ff5733"
print(Unsure("Order Total: 12345 USD").flat_map_to("price")) # "12345"Transforms the string into what's demanded but returns an Unsure, so it's chainable. Example:
from unsure_ai import Unsure, config_global_unsure
config_global_unsure(groq_api_key='your key here')
print(Unsure("Amount: 200.23 $").map_to("number").map_to("integer").flat()) # "200"Returns either the changed transformation's result or the initial value if no mapTo was used. Example:
from unsure_ai import Unsure, config_global_unsure
config_global_unsure(groq_api_key='your key here')
print(Unsure("Amount: 200.23 $").map_to("number").map_to("integer").flat()) # "200"
print(Unsure("Some value").flat()) # "Some value"inferenceEndpoint: The function that will be called in the operators.
groqApiKey: The Api key that will be used to call groq APIs using llama3-70b-8192 model.
openAiApiKey: The Api key that will be used to call Open Ai APIs using gpt-3.5-turbo.
model: You can specify the model you want to use, for open source models llama3-70b-8192 works best which is the default.
preventLowerCase: Prevents lowercasing the inference response.
Note: If both inferenceEndpoint and groqApiKey are provided inferenceEndpoint will be used.
You might have noticed so far that a global Unsure instance is used. You can also create your own instance with it's own configuration.
from unsure_ai import Unsure, create_unsure, config_global_unsure
config_global_unsure(groq_api_key='your key here')
my_unsure = create_unsure(openai_api_key='your key here')
print(Unsure("Amount: 200.23 $").map_to("number").map_to("integer").flat()) # Uses Groq
print(my_unsure("Some value").flat()) # Uses OpenAI APIsThis project is under the ISC license. Requests and contributions are most welcomed.
