Using Flask

README.md

@@ -1,5 +1,5 @@
 ---
-title: Live Transcription
+title: Utilities
 colorFrom: blue
 colorTo: blue
 sdk: gradio

app.py

@@ -1,253 +1,88 @@
-from authlib.integrations.flask_client import OAuth
-from authlib.common.security import generate_token
-import ffmpeg
-from flask import Flask, render_template, request, jsonify, url_for, redirect, session
-from functools import wraps
-import os
-import streamlink
-import threading
-import time
-from faster_whisper import WhisperModel
-import subprocess
+from flask import Flask, render_template, request, jsonify
+from qdrant_client import QdrantClient
+from qdrant_client import models
+import torch.nn.functional as F
+import torch
+from torch import Tensor
+from transformers import AutoTokenizer, AutoModel
+from qdrant_client.models import Batch, PointStruct
+from pickle import load, dump
+import numpy as np
+import os, time, sys
 from datetime import datetime as dt
-from datetime import timedelta, timezone
-from apiclient import discovery
-from google.oauth2 import service_account
-import json
-
-# Import secrets
-client_secret = json.loads(os.environ.get("client_secret"))
-gdoc_id = os.environ.get("gdoc_id")
-GOOGLE_CLIENT_SECRET = os.environ.get("GOOGLE_CLIENT_SECRET")
-GOOGLE_CLIENT_ID = os.environ.get("GOOGLE_CLIENT_ID")
-allowed_users = os.environ.get("allowed_users")
-
-# Faster Whisper setup
-model_size = 'small'
-beamsize = 2
-wmodel = WhisperModel(model_size, device="cpu", compute_type="int8")
-
-# Delete local_transcript if it exists
-if not os.path.exists('transcription_files'): os.makedirs('transcription_files')
-for f in os.listdir('transcription_files/'): os.remove(os.path.join('transcription_files/', f)) # clear any old files in transcription_files folder
-
-with open("client_secret.json", "w") as json_file: json.dump(client_secret, json_file, indent=4)
-
-scopes = ["https://www.googleapis.com/auth/documents", "https://www.googleapis.com/auth/drive.file"]
-credentials = service_account.Credentials.from_service_account_file('client_secret.json', scopes=scopes)
-service = discovery.build('docs', 'v1', credentials=credentials)
-
-local_tz = 5.5 # For timestamps
-local_transcript = 'transcription_files/tr.txt'
-pid_file = 'transcription_files/pid.txt'
-
-# Check if mp3 folder exists, and create it if it doesn't
-if not os.path.exists('mp3'): os.makedirs('mp3')
-# Delete any old files in mp3 folder
-for f in os.listdir('mp3/'): os.remove(os.path.join('mp3/', f))
-
-app = Flask(__name__, static_url_path='/static')
-app.secret_key = os.urandom(12)
-
-oauth = OAuth(app)
-
-# Store the streamlink process
-stream_process = None
-recording = False
-mp3_extraction_process = None
-
-def update_gdoc(text, gdoc_id): # Update contents of google doc
-    print('Updating Google Doc', gdoc_id)
-    doc = service.documents().get(documentId=gdoc_id).execute()
-    endindex = [p['endIndex'] for p in doc['body']['content'] if 'paragraph' in p][-1]
-
-    try: 
-        body = {'requests': [{'insertText': {'location': {'index': endindex-1,}, 'text': ' ' + text}}]}
-        result = service.documents().batchUpdate(documentId=gdoc_id, body=body).execute()
-        print(result)
-
-    except Exception as e:
-        print(e)
-
-def process_complete_callback(retcode, **kwargs):
-    if retcode == 0:
-        print("FFmpeg process completed successfully!")
-    else:
-        print("FFmpeg process encountered an error.")
-
-def transcribe_audio(latest_file, time_counter):
-    print('transcribing ', latest_file)
-    segments, info = wmodel.transcribe(f"{latest_file}", beam_size=beamsize) # beamsize is 2.
-    text = ''
-
-    for segment in segments:
-        text += segment.text
-    transcribed = text.replace('\n', ' ').replace('  ', ' ')
-    if time_counter%5  == 0: 
-        transcribed_sents = transcribed.split('. ') # Get the first fullstop break and append to previous para, before adding time code
-        transcribed = transcribed_sents[0] + '\nTime ' + str((dt.now(timezone.utc) + timedelta(hours=local_tz)).strftime('%H:%M:%S')) + '\n' + '. '.join(transcribed_sents[1:])
-
-    time_counter += 1
-    return transcribed, time_counter
-
-def save_audio(youtube_url):
-    global stream_process, recording, mp3_extraction_process
-    try:
-        streams = streamlink.streams(youtube_url)
-        #if "audio" not in streams:
-        #    raise Exception("No audio stream found.")
-        
-        stream_url = streams["144p"].url
-        time_counter = 0
-        while recording:
-            # Save audio only into mp3 files
-
-            saved_mp3 = f"mp3/audio_{int(time.time())}.mp3"
-            mp3_extraction_process = (
-                ffmpeg
-                .input(stream_url, t=30)
-                .audio
-                # TODO - change destination url to relevant url
-                .output(saved_mp3)
-                .overwrite_output()
-                .global_args('-loglevel', 'panic')
-                .run_async()
-                )
-            
-            print('pid', mp3_extraction_process.pid)
-            # write the pid to pid_file
-            with open(pid_file, 'w') as f: f.write(str(mp3_extraction_process.pid))
-            
-            # If there is more than one mp3 file in the folder, transcribe the one that is not being written to
-            mp3files = [f for f in os.listdir('mp3') if f.endswith('.mp3')]
-            if len(mp3files) < 2: 
-                print('Sleeping for 30s as only one mp3 file in folder')
-                time.sleep(30)            
-            else: 
-                starttime = time.time()
-                file_to_transcribe = [f for f in mp3files if f != os.path.basename(saved_mp3)][0]
-                print('Working on ', file_to_transcribe)
-                transcribed, time_counter = transcribe_audio(f'mp3/{file_to_transcribe}', time_counter)
-                os.remove(f'mp3/{file_to_transcribe}')
-
-                update_gdoc(transcribed, gdoc_id)
-                with open(local_transcript, 'a', encoding='utf-8', errors='ignore') as f: f.write(transcribed)
-                
-                elapsed_time = time.time() - starttime
-                print('Time to transcribe:', elapsed_time, 'seconds')
-                if elapsed_time < 30: 
-                    print(f'Sleeping for {30-elapsed_time} as there are more than one mp3 files in folder')
-                    time.sleep(30-elapsed_time)
-            #time.sleep(30)
-
-    except Exception as e:
-        recording = False
-        print('exception', str(e))
-        return str(e)
-
-@app.route("/start_process", methods=["POST"])
-def start_process():
-    if not os.path.isfile(local_transcript):
-        global recording, stream_process
-        with open(local_transcript, 'a', encoding='utf-8', errors='ignore') as f: f.write('') # Create the local transcript file, which is used as a check to prevent multiple recordings
-
-        youtube_url = request.form.get("url")
-        if not youtube_url:
-            return jsonify({"message": "Please provide a valid YouTube URL."}), 400
-
-        if recording:
-            return jsonify({"message": "A recording is already in progress."}), 400
-        
-        print('In start process')
-        recording = True
-        stream_process = threading.Thread(target=save_audio, args=(youtube_url,))
-        stream_process.start()
-
-        return jsonify({"message": "Recording started."}), 200
+from datetime import timedelta
+from datetime import timezone
+
+app = Flask(__name__)
+
+# Initialize Qdrant Client and other required settings
+qdrant_api_key = os.environ.get("qdrant_api_key")
+qdrant_url = os.environ.get("qdrant_url")
+
+client = QdrantClient(url=qdrant_url, port=443, api_key=qdrant_api_key, prefer_grpc=False)
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+def average_pool(last_hidden_states: Tensor,
+                 attention_mask: Tensor) -> Tensor:
+    last_hidden = last_hidden_states.masked_fill(~attention_mask[..., None].bool(), 0.0)
+    return last_hidden.sum(dim=1) / attention_mask.sum(dim=1)[..., None]
+
+tokenizer = AutoTokenizer.from_pretrained('intfloat/e5-base-v2')
+model = AutoModel.from_pretrained('intfloat/e5-base-v2').to(device)
+
+def e5embed(query):
+  batch_dict = tokenizer(query, max_length=512, padding=True, truncation=True, return_tensors='pt')
+  batch_dict = {k: v.to(device) for k, v in batch_dict.items()}
+  outputs = model(**batch_dict)
+  embeddings = average_pool(outputs.last_hidden_state, batch_dict['attention_mask'])
+  embeddings = F.normalize(embeddings, p=2, dim=1)
+  embeddings = embeddings.cpu().detach().numpy().flatten().tolist()
+  return embeddings
+
+@app.route("/")
+def index():
+    return render_template("index.html")
+
+@app.route("/search", methods=["POST"])
+def search():
+    query = request.form["query"]
+    topN = 200  # Define your topN value
+
+
+    print('QUERY: ',query)
+    if query.strip().startswith('tilc:'):
+        collection_name = 'tils'
+        qvector = "context"
+        query = query.replace('tilc:', '')
+    elif query.strip().startswith('til:'):
+        collection_name = 'tils'
+        qvector = "title"
+        query = query.replace('til:', '')
+    else: collection_name = 'jks'
+
+    timh = time.time()
+    sq = e5embed(query)    
+    print('EMBEDDING TIME: ', time.time() - timh)
+
+    timh = time.time()
+    if collection_name == "jks": results = client.search(collection_name=collection_name, query_vector=sq, with_payload=True, limit=topN)
+    else: results = client.search(collection_name=collection_name, query_vector=(qvector, sq), with_payload=True, limit=100)
+    print('SEARCH TIME: ', time.time() - timh)
     
-    else: return jsonify({"message": "Recording is already in progress."}), 400
-
-
-@app.route("/stop_process", methods=["POST"])
-def stop_process():
-    global recording, stream_process, mp3_extraction_process
-
-    if not recording:
-        return jsonify({"message": "No recording is currently in progress."}), 400
-    print('In stop process')
-    recording = False
-    stream_process.join()
-    stream_process = None
-    mp3_extraction_process.terminate()
-    mp3_extraction_process = None
-    for f in os.listdir('mp3/'): os.remove(os.path.join('mp3/', f))
-    if os.path.isfile(local_transcript): os.remove(local_transcript)
-    # check if pid_file exists, get the pid inside it and convert to int, and use os.kill to kill it
-    if os.path.isfile(pid_file): 
-        with open(pid_file, 'r') as f: pid = int(f.read())
-        try: 
-            os.kill(pid, 9) # For linux
-            print("Process terminated successfully in Linux.")
-        except: 
-            try: 
-                process = subprocess.Popen(["taskkill", "/F", "/PID", str(pid)], stdout=subprocess.PIPE, stderr=subprocess.PIPE) # For Windows
-                process.communicate()
-                print("Process terminated successfully in Windows.")
-            except Exception as e:
-                print("Error:", e) 
-        os.remove(pid_file)
-
-    return jsonify({"message": "Recording stopped."}), 200
-
-@app.route('/google/')
-def google():
-    CONF_URL = 'https://accounts.google.com/.well-known/openid-configuration'
-    oauth.register(
-        name='google',
-        client_id=GOOGLE_CLIENT_ID,
-        client_secret=GOOGLE_CLIENT_SECRET,
-        server_metadata_url=CONF_URL,
-        client_kwargs={"scope": "openid email profile"}
-    )
-
-    # Redirect to google_auth function/page
-    redirect_uri = url_for('google_auth', _external=True)
-    session['nonce'] = generate_token()
-    return oauth.google.authorize_redirect(redirect_uri, nonce=session['nonce'])
-
-@app.route('/google/auth/')
-def google_auth():
-    token = oauth.google.authorize_access_token()
-    user = oauth.google.parse_id_token(token, nonce=session['nonce'])
-    session['user'] = user
-    print('USER', user)
-    # Redirect to home if login successful
-    return redirect('/home')
-
-def is_not_logged_in():
-    return session.get('user') is None or session.get('nonce') is None
-
-# decorator to check if user is logged in, used for protected URLs
-def login_required(f):
-    @wraps(f)
-    def decorated_function(*args, **kwargs):
-        if is_not_logged_in():
-            return redirect('/login')
-        return f(*args, **kwargs)
-    return decorated_function
-
-@app.route("/home")
-@login_required
-def home():
-    return render_template("home.html")
-
-@app.route("/", methods=["GET"])
-@app.route("/login", methods=["GET"])
-def login():
-    if not is_not_logged_in():
-        return redirect("/home")
-    #return render_template("login.html")
-    return render_template("home.html")
+    print(results[0].payload['text'].split('\n'))
+    try: 
+        results = [{"text": x.payload['text'], "date": str(int(x.payload['date'])), "id": x.id} for x in results]  # Implement your Qdrant search here     
+        return jsonify(results)
+    except:
+        return jsonify([])
+
+@app.route("/delete_joke", methods=["POST"])
+def delete_joke():
+    joke_id = request.form["id"]
+    print('Deleting joke no', joke_id)
+    client.delete(collection_name="jks", points_selector=models.PointIdsList(points=[int(joke_id)],),)
+    return jsonify({"deleted": True})
 
 if __name__ == "__main__":
     app.run(host="0.0.0.0", debug=True, port=7860)

modules/dataset.py

@@ -1,19 +0,0 @@
-from datasets import load_dataset
-
-dataset = load_dataset("go_emotions", split="train")
-
-emotions = dataset.info.features['labels'].feature.names
-
-def query_emotion(start, end):
-    rows = dataset[start:end]
-    texts, labels = [rows[k] for k in rows.keys()]
-
-    observations = []
-
-    for i, text in enumerate(texts):
-        observations.append({
-            "text": text,
-            "emotion": emotions[labels[i]],
-        })
-
-    return observations

modules/inference.py

@@ -1,11 +0,0 @@
-from transformers import T5Tokenizer, T5ForConditionalGeneration
-
-tokenizer = T5Tokenizer.from_pretrained("t5-small")
-model = T5ForConditionalGeneration.from_pretrained("t5-small")
-
-
-def infer_t5(input):
-    input_ids = tokenizer(input, return_tensors="pt").input_ids
-    outputs = model.generate(input_ids)
-
-    return tokenizer.decode(outputs[0], skip_special_tokens=True)

requirements.txt

@@ -1,9 +1,5 @@
-google-api-python-client
-google-auth-oauthlib
-streamlink
-faster-whisper
-requests
-ffmpeg-python
-Authlib
+transformers
+torch
+qdrant-client
 flask
-Werkzeug
+Werkzeug

static/script.js

@@ -1,47 +0,0 @@
-$(document).ready(function () {
-  let recording = false;
-
-  $("#startBtn").click(function () {
-    const youtubeUrl = $("#urlInput").val().trim();
-    if (youtubeUrl === "") {
-      showMessage("Please enter a valid YouTube Livestream URL.", "danger");
-      return;
-    }
-
-    // Call the start_process route in Flask
-    $.ajax({
-      type: "POST",
-      url: "/start_process",
-      data: { url: youtubeUrl },
-      success: function (data) {
-        showMessage(data.message, "success");
-        recording = true;
-      },
-      error: function (xhr, status, error) {
-        showMessage("Error: " + xhr.responseText, "danger");
-      },
-    });
-  });
-
-  $("#stopBtn").click(function () {
-    showMessage("Stopping transcription. This may take upto 30 sec.", "success");
-    // Call the stop_process route in Flask
-    $.ajax({
-      type: "POST",
-      url: "/stop_process",
-      success: function (data) {
-        showMessage(data.message, "success");
-        recording = false;
-      },
-      error: function (xhr, status, error) {
-        showMessage("Error: " + xhr.responseText, "danger");
-      },
-    });
-  });
-
-  function showMessage(message, type) {
-    $("#message").html(
-      `<div class="alert alert-${type}" role="alert">${message}</div>`
-    );
-  }
-});

static/style.css

@@ -1,37 +0,0 @@
-#container {
-  width: 600px;
-  margin: 0 auto;
-  text-align: center;
-}
-
-#urlInput {
-  width: 500px;
-}
-
-button {
-  margin: 10px;
-}
-
-#login {
-  display: flex;
-  justify-content: center;
-  flex-direction: row;
-}
-
-#loginButton {
-  display: flex;
-  background-color: rgb(255, 255, 255);
-  justify-content: center;
-  align-items: center;
-  width: fit-content;
-  padding: 10px 30px;
-  box-shadow: 1px 1px 1px 1px rgb(170, 170, 170);
-  border-radius: 15px;
-  border: 1px solid rgb(170, 170, 170);
-}
-
-#loginButtonText {
-  justify-content: center;
-  padding: 10px;
-  color: rgb(85, 85, 85);
-}

templates/home.html

@@ -1,85 +0,0 @@
-<!DOCTYPE html>
-<html>
-
-<head>
-  <title>YouTube Livestream Audio Recorder</title>
-  <!-- Add necessary CSS and jQuery libraries -->
-  <style>
-    body {
-            font-family: Roboto, sans-serif;
-            margin: 0;
-            padding: 0;
-			height: 100%;
-    }
-
-    #container {
-      width: 600px;
-      margin: 0 auto;
-      text-align: center;
-    }
-
-    #url {
-      width: 500px;
-    }
-
-    button {
-      margin: 10px;
-    }
-
-    .row, h2 {
-            display: flex;
-            align-items: center;
-            margin: 10px;
-    }
-
-    select, button {
-            padding: 5px;
-            border: 1px solid #ccc;
-            border-radius: 5px;
-    }
-
-    input {
-            padding: 5px;
-            border: 1px solid #ccc;
-            border-radius: 5px;
-            width: 500px;
-    }
-
-    label, #message {
-            margin: 0px 5px 0px 10px;
-    }
-
-    button {
-            background-color: #2196f3;
-            color: white;
-            cursor: pointer;
-            margin: 0px 0px 0px 3px;
-    }
-
-        button:hover {
-            background-color: #1976d2;
-    }
-  </style>
-</head>
-
-<body>
-  <div class="container">
-    <h2>Search & Transcribing Utilities</h2>
-    <div class="mb-3 row">
-      <label for="urlInput" class="form-label">Enter YouTube Livestream URL:</label>
-      <input type="text" class="form-control" id="urlInput"
-        placeholder="e.g., https://www.youtube.com/watch?v=YOUR_STREAM_ID">
-    
-    <button class="btn btn-primary" id="startBtn">Start</button>
-    <button class="btn btn-danger" id="stopBtn">Stop</button>
-    <div id="message"></div>
-  </div>
-
-  <!-- Add necessary jQuery library -->
-  <script src="https://cdnjs.cloudflare.com/ajax/libs/jquery/3.6.0/jquery.min.js"></script>
-  <!-- Add custom script for handling button clicks -->
-  <script src="{{ url_for('static', filename='script.js') }}"></script>
-
-</body>
-
-</html>

templates/index.html

@@ -1,85 +1,112 @@
 <!DOCTYPE html>
-<html>
-
+<html lang="en">
 <head>
-  <title>YouTube Livestream Audio Recorder</title>
-  <!-- Add necessary CSS and jQuery libraries -->
-  <style>
-    body {
-            font-family: Roboto, sans-serif;
-            margin: 0;
-            padding: 0;
-			height: 100%;
-    }
-
-    #container {
-      width: 600px;
-      margin: 0 auto;
-      text-align: center;
-    }
-
-    #url {
-      width: 500px;
-    }
-
-    button {
-      margin: 10px;
-    }
-
-    .row, h2 {
+    <meta charset="UTF-8">
+    <title>Joke Search</title>
+    <script src="https://code.jquery.com/jquery-3.6.0.min.js"></script>
+    <style>
+        body {
+            font-family: Arial, sans-serif;
+        }
+
+        h1 {
+            text-align: center;
+        }
+
+        #search-container {
             display: flex;
+            justify-content: center;
             align-items: center;
-            margin: 10px;
-    }
+        }
 
-    select, button {
-            padding: 5px;
-            border: 1px solid #ccc;
-            border-radius: 5px;
-    }
+        #query {
+            width: 70%;
+            padding: 10px;
+        }
 
-    input {
-            padding: 5px;
-            border: 1px solid #ccc;
-            border-radius: 5px;
-            width: 500px;
-    }
+        #search {
+            background-color: #4c72af;
+            color: white;
+            border: none;
+            padding: 10px 20px;
+            cursor: pointer;
+            margin-left: 10px; /* Add margin to separate the query bar and button */
+        }
 
-    label, #message {
-            margin: 0px 5px 0px 10px;
-    }
+        #results {
+            margin: 20px;
+        }
 
-    button {
-            background-color: #2196f3;
+        .result {
+            border: 1px solid #ccc;
+            padding: 10px;
+            margin: 10px 0;
+        }
+
+        .delete {
+            background-color: #ff0000;
             color: white;
+            border: none;
+            padding: 5px 10px;
             cursor: pointer;
-            margin: 0px 0px 0px 3px;
-    }
-
-        button:hover {
-            background-color: #1976d2;
-    }
-  </style>
+            margin: 0 10px;
+        }
+    </style>
 </head>
-
 <body>
-  <div class="container">
-    <h2>Search & Transcribing Utilities</h2>
-    <div class="mb-3 row">
-      <label for="urlInput" class="form-label">Enter YouTube Livestream URL:</label>
-      <input type="text" class="form-control" id="urlInput"
-        placeholder="e.g., https://www.youtube.com/watch?v=YOUR_STREAM_ID">
-    
-    <button class="btn btn-primary" id="startBtn">Start</button>
-    <button class="btn btn-danger" id="stopBtn">Stop</button>
-    <div id="message"></div>
-  </div>
-
-  <!-- Add necessary jQuery library -->
-  <script src="https://cdnjs.cloudflare.com/ajax/libs/jquery/3.6.0/jquery.min.js"></script>
-  <!-- Add custom script for handling button clicks -->
-  <script src="{{ url_for('static', filename='script.js') }}"></script>
+    <h1>Joke Search</h1>
+    <div id="search-container">
+        <input type="text" id="query" placeholder="Search...">
+        <button id="search">Search</button>
+    </div>
+    <div id="results"></div>
 
-</body>
+    <script>
+        $(document).ready(function () {
+            function performSearch() {
+                var query = $("#query").val();
+
+                $.post("/search", { query: query }, function (data) {
+                    var results = $("#results");
+                    results.empty();
+
+                    data.forEach(function (result) {
+                        var resultElement = $("<div class='result'>" +
+                            "<p>" + result.text + "</p>" +
+                            "<small>Date: " + result.date + "</small>" +
+                            "<button class='delete' data-id='" + result.id + "'>Delete</button>" +
+                            "</div>");
+
+                        results.append(resultElement);
 
+                        resultElement.find(".delete").on("click", function () {
+                            var id = $(this).data("id");
+                            var resultButton = $(this);
+
+                            $.post("/delete_joke", { id: id }, function (data) {
+                                // Handle the delete response if needed
+
+                                if (data.deleted) {
+                                    // Replace the button with "DELETED!" text
+                                    resultButton.replaceWith("<p class='deleted-text'>DELETED!</p>");
+                                }
+                            });
+                        });
+                    });
+                });
+            }
+
+            $("#search").on("click", function () {
+                performSearch();
+            });
+
+            $("#query").on("keydown", function (event) {
+                if (event.key === "Enter") {
+                    event.preventDefault(); // Prevent form submission
+                    performSearch();
+                }
+            });
+        });
+    </script>
+</body>
 </html>

templates/login.html

@@ -1,27 +0,0 @@
-<!DOCTYPE html>
-<html lang="en">
-
-<head>
-    <meta charset="UTF-8">
-    <meta name="viewport" content="width=device-width, initial-scale=1.0">
-    <link rel="preconnect" href="https://fonts.googleapis.com">
-    <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
-    <link href="https://fonts.googleapis.com/css2?family=Poppins&display=swap" rel="stylesheet">
-    <link rel="stylesheet" type="text/css" href="{{ url_for('static',filename='style.css') }}">
-    <title>Flask app</title>
-</head>
-
-<body>
-    <div id="login">
-        <a href="google/" id="loginButton" style="text-decoration:none;">
-            <img id="google"
-                src="https://upload.wikimedia.org/wikipedia/commons/thumb/5/53/Google_%22G%22_Logo.svg/2048px-Google_%22G%22_Logo.svg.png"
-                alt="Google" style="width:50px;height:50px;">
-            <div id="loginButtonText">
-                Login with Google
-            </div>
-        </a>
-    </div>
-</body>
-
-</html>