experiments.py

from chromadb.utils import embedding_functions

from embedding.chroma import ChromaDB


def add_docs_to_db(doctxt, id):
    """_summary_

    Args:
        doctxt (_type_): _description_
        id (_type_): _description_

    Raises:
        FileNotFoundError: _description_
        FileNotFoundError: _description_

    Returns:
        _type_: _description_
    """
    db = ChromaDB()
    dev = db.medical_clerking
    dev.add(documents=doctxt, ids=id)
    print("documents added to database successfully")
    return


class Embedder:
    def __init__(self):
        self.default_ef = embedding_functions.DefaultEmbeddingFunction()

    def create_embeddings(self, userinput):
        """create embeddings

        Args:
            userinput (string): text data

        Raises:
            FileNotFoundError: _description_
            FileNotFoundError: _description_

        Returns:
            _type_: _description_
        """
        input = [userinput]
        val = self.default_ef(input)
        print(val)
        return val

# Example usage:


embedder = Embedder()
eg = "Doctor I have a pneumonia"
eg2 = "oh my leg hurts"
v2 = embedder.create_embeddings(userinput=eg)
v3 = embedder.create_embeddings(userinput=eg2)
# model architecture
# data loader (will load the input data)
# data processor (will clean the loaded data)
# embeddings generator (will then create an embedding)
# feature suggestion (let the user select other llm)
# nearest distance calculator
# feature suggestion (the user can select top 3 vs 5 vs 10)
# serve the nearest distances

# from pathlib import Path

# import pandas as pd
# import torch
# from transformers import AutoModel, AutoTokenizer


# # data loader (will load the input data)
# def dataloader(path):
#     """
#     load csv file
#     """

#     try:
#         path = Path(path)

#         if path.is_dir():
#             csv_files = list(path.glob("*.csv"))
#             if not csv_files:
#                 raise FileNotFoundError(f"No csv files found in directory: {path}")
#             print(f" Found {len(csv_files)} CSV Files(s) in this directory")
#             pd.read_csv(csv_files[0])
#             print(f"Successfully loaded: {csv_files[0].name}")

#         elif path.suffix == ".csv":
#             if not path.exists():
#                 raise FileNotFoundError()
#             print(f"File found at {path}")
#             pd.read_csv(path)
#     except Exception as e:
#         print(f"Error found {e}")


# # data processor (will clean the loaded data)

# data = {
#     'x': ["Doctor, I have a shortness of breath and chest pain and also have temperatures",
#           "My leg hurts", "I am feeling dizzy"],
#     'y': ["pneumonia", "dvt", "vertigo"]
# }
# df = pd.DataFrame(data)

# tokenizer = AutoTokenizer.from_pretrained("Charangan/MedBERT")
# model = AutoModel.from_pretrained("Charangan/MedBERT")


# def text_to_embedding(text):
#     # Tokenize the text
#     inputs = tokenizer(text, return_tensors='pt', truncation=True, padding=True)
#     # Get the embeddings
#     with torch.no_grad():
#         outputs = model(**inputs)
#     # The embeddings are in the last hidden state
#     embeddings = outputs.last_hidden_state
#     # You can use the mean of the embeddings as the representation
#     mean_embeddings = embeddings.mean(dim=1)
#     return mean_embeddings

# # Apply the function to the 'x' column


# df['x_embeddings'] = df['x'].apply(lambda x: text_to_embedding(x).squeeze().numpy())