Different LLMs with funcchain EASY TO USE

Example

See llamacpp.py Also see supported MODELS.md

In this example, we will use the funcchain library to perform sentiment analysis on a piece of text. This showcases how funcchain can seamlessly utilize different Language Models (LLMs) using local llamacpp models, without many code changes..

This is particularly useful for developers looking to integrate different models in a single application or just experimenting with different models.

Full Code Example

from funcchain import chain, settings
from pydantic import BaseModel, Field
from rich import print

# define your model
class SentimentAnalysis(BaseModel):
    analysis: str = Field(description="A description of the analysis")
    sentiment: bool = Field(description="True for Happy, False for Sad")

# define your prompt
def analyze(text: str) -> SentimentAnalysis:
    """
    Determines the sentiment of the text.
    """
    return chain()

# set global llm
settings.llm = "llamacpp/openchat-3.5-0106:Q3_K_M"

# log tokens as stream to console
settings.console_stream = True

# run prompt
poem = analyze("I really like when my dog does a trick!")

# show final parsed output
print(poem)

Demo

poem = analyze("I really like when my dog does a trick!")

$ {"analysis": "A dog trick", "sentiment": true}

SentimentAnalysis(analysis='A dog trick', sentiment=True)

Instructions

Step-by-step

Necessary Imports

from funcchain import chain, settings
from pydantic import BaseModel, Field

Define the Data Model

Here, we define a SentimentAnalysis model with a description of the sentiment analysis and a boolean field indicating the sentiment.

class SentimentAnalysis(BaseModel):
    analysis: str = Field(description="A description of the analysis")
    sentiment: bool = Field(description="True for Happy, False for Sad")

Create the Analysis Function

This 'analyze' function takes a string as input and is expected to return a SentimentAnalysis object by calling the chain() function from the funcchain library.

def analyze(text: str) -> SentimentAnalysis:
    """
    Determines the sentiment of the text.
    """
    return chain()

Execution Configuration

In the main block, configure the global settings to set the preferred LLM, enable console streaming, and run the analyze function with sample text.

# set global llm
settings.llm = "llamacpp/openchat-3.5-0106:Q3_K_M"

# log tokens as stream to console
settings.console_stream = True

# run prompt
poem = analyze("I really like when my dog does a trick!")

# show final parsed output
print(poem)

Important

We need to note here is that settings.llm can be adjusted to any model mentioned in MODELS.md and your funcchain code will still work and chain() does everything in the background for you.