z_generate.py

from huggingface_hub import InferenceClient
import os 

class ServerlessInference:
    def __init__(self, vector_store_text = None, vector_store_images = None):
        self.model:str = "HuggingFaceH4/zephyr-7b-beta"
        self.client = InferenceClient(api_key=os.getenv("HF_SERVELESS_API"))
        self.vs_text = vector_store_text
        self.vs_images = vector_store_images

    def test(self, query:str) -> str:
        '''Responds to query using llm'''
        messages:str = [
            {
                "role": "user",
                "content": query
            }
        ]
        completion = self.client.chat.completions.create(
            model=self.model,
            messages=messages, 
            max_tokens=500
        )

        return completion.choices[0].message.content
    
    def perform_rag(self, query:str):
        # First perform text search 
        relevant_docs = self.vs_text.similarity_search(query=query, k=5)
        return relevant_docs[0].page_content