app.py

import streamlit as st
import pandas as pd
from huggingface_hub import Repository
import os 
from pathlib import Path
import json
import numpy as np

 
# Declaring the variables for later use to talk to dataset

# the token is saved as secret key-value pair in the environment which can be access as shown below
auth_token = os.environ.get("space_to_dataset") or True

DATASET_REPO_URL = 'ppsingh/annotation_data'   # path to dataset repo
DATA_FILENAME = "paralist.json"
DATA_FILE = os.path.join("data", DATA_FILENAME)

# cloning the dataset repo


# Data file name
file_name = 'paralist.json'

# reading the json
@st.cache(allow_output_mutation=True)
def read_dataset():
    repo = Repository( local_dir="data", clone_from=DATASET_REPO_URL, repo_type="dataset", use_auth_token= auth_token)
    with open('data/{}'.format(file_name), 'r', encoding="utf8") as json_file:
      paraList = json.load(json_file)
    
    return repo, paraList

st.sidebar.markdown(""" 
   # Data Annotation Demo 
This app is demo how to use the space to provide user interface for the data annotation/tagging. The data resides in repo_type 'dataset'.
""")
# sidebar with info and drop down to select from the keys

topic = None
repo, paraList = read_dataset()
# getting outer level keys in json 
keys = paraList.keys()  

if keys is not None:
  topic = st.sidebar.selectbox(label="Choose dataset topic to load", options=keys )

 
#with st.container():

      
with st.form("annotation_form"):
    if topic is not None:
        subtopics = list(paraList[topic].keys())
  #st.write(subtopics)
    val = np.random.randint(0,len(subtopics)-1)
    tag = subtopics[val]
  
    idx = np.random.randint(0,3)
    
    st.markdown("**Text**")
    st.write(paraList[topic][tag][idx]['textsegment'])
    
    st.markdown("**Tag**")
    st.write(tag)
    
    feedback = st.selectbox('0 If Tag is not a good keyword for text, 5 for prefect match',(0,1,2,3,4,5))
    submitted = st.form_submit_button("Submit")
    if submitted:
        paraList[topic][tag][idx]['annotation'].append(feedback)
        with open("data/{}".format(file_name), "w") as outfile:
            json.dump(paraList, outfile)
        repo.push_to_hub('added new annotation')
              # st.write(type(paraList))
    
      
      #c1, c2, c3 = st.columns([3, 1, 1])
      #with c1:
       #   st.header('Text')
        #  st.write(paraList[topic][tag][idx]['textsegment'])
  
      #with c2:
       #   st.header('Tag')
        #  st.text(tag)
  
      #with c3:
       #   st.header('Feedback')
        #  feedback = None
         # feedback = st.selectbox('0 If Tag is not a good keyword for text, 5 for prefect match',(0,1,2,3,4,5)) 
          #if feedback:
           #   st.write(feedback)
#      if st.button('Submit'):
#        paraList[topic][choice][idx]['annotation'].append(feedback)
#      with open('data/{}'.format(file_name), 'r', encoding="utf8") as json_file:
 #       json.dump(paraList,json_file, ensure_ascii = True)
  #      repo.push_to_hub('added new annotation')
        
#st.write(paraList)      
    #new_row  = title
#  data = data.append(new_row, ignore_index=True)
#  st.write(data)
#  st.write(os.getcwd())
#  data.to_csv('test.csv', index= False)


#st.write(df)
#   st.write('data/test.csv')
# iterate over files in
# that directory        
#directory = os.getcwd()
#files = Path(directory).glob('*')
#for file in files:
#    st.write(file)

#with open(DATA_FILE, "a") as csvfile:
#  writer = csv.DictWriter(csvfile, fieldnames=["Sentences"])
#  writer.writerow({'Sentences': new_row})
#  repo.push_to_hub('adding new line')
#  st.write('Succcess')