LLM Functions ⚙️

Easily define custom LLM (Language Learning Model) functions with structured input and output.

Getting Started

1. Installation

Install the package using pip:

pip install llm-functions

2. Setting Up Your First Function

Let's create a sentiment classification function:

Step 1: Create a directory named sentiment. Inside this directory, you need two files:

└─ sentiment
   ├── function_args.json
   ├── model_args.json
   └── template.txt

Step 2: Define your prompt template in template.txt:

Aalyze and determine whether the sentiment of the following sentence is positive, negative, or neutral.
Sentence:
{sentence}

Step 3: Set model paramters in model_args.json:

{
    "temperature": 0.0,
    "model": "gpt-3.5-turbo"
}

Step 4: Specify the function's output format in function_args.json:

{
    "function_name": "sentiment_classifier",
    "description": "Classify the sentiment of a sentence",
    "properties": {
        "thoughts": {
            "type": "string",
            "description": "Your thoughts when classifying sentiment of the given sentence."
        },
        "sentiment": {
            "type": "string",
            "enum": ["positive", "negative", "neutral"],
            "description": "The sentiment of the given sentence."
        }
    }
}

3. Using Your LLM Function in Python

Here's how you can load and use your LLM function:

from llm_functions import LLMFunction

# Load your function
sentiment = LLMFunction.from_dir("./sentiment/")

# Make a prediction
pred = sentiment(sentence="I am super happy!")

# Expected Output:
# {
#   "rationale": "The sentiment in the sentence is positive",
#   "sentiment": "positive"
# }