utils.py

import chromadb
from sentence_transformers import CrossEncoder, SentenceTransformer
import json

def chroma_client_setup():
    chroma_client = chromadb.Client()
    collection = client.create_collection(
        name="food_collection",
        metadata={"hnsw:space": "cosine"} # l2 is the default
    )
    return collection

def load_data():
    with open("test_json.json", "r") as f:
        data = json.load(f)
    return data

def embedding_function(items_to_embed: list[str]):
    sentence_model = SentenceTransformer(
        "mixedbread-ai/mxbai-embed-large-v1"
    )
    embedded_items = sentence_model.encode(
        items_to_embed,
        show_progress_bar=True
    )
    return embedded_items

def chroma_upserting(collection, payload:list[dict]):
    collection.add(
        documents=[item['doc'] for item in payload],
        embeddings=embedding_function([item['doc'] for item in payload]),
        metadatas=payload,
        ids=[f"id{item}" for item in range(len(embedfings))]
    )

def search_chroma(collection, query:str):
    results = collection.query(
        query_embeddings=embedding_function([query]),
        n_results=5
    )
    return results

def reranking_results(query: str, top_k_results: list[str]):
    # Load the model, here we use our base sized model
    rerank_model = CrossEncoder("mixedbread-ai/mxbai-rerank-xsmall-v1")
    reranked_results = rerank_model.rank(query, top_k_results, return_documents=True)
    return reranked_results