initial commit

.gitignore

@@ -1,3 +1,6 @@
+.idea
+venv
+
 # ---> Python
 # Byte-compiled / optimized / DLL files
 __pycache__/

README.md

@@ -1,3 +1,3 @@
 # summarizer-website
 
-A simple website to summarize text using AI.
+_A simple website to summarize text using AI._

requirements.txt

@@ -0,0 +1,4 @@
+openai
+flask
+tiktoken
+Flask-CORS

server.py

@@ -0,0 +1,147 @@
+import os
+from urllib.parse import urlparse
+
+import openai
+import tiktoken
+from flask import Flask, render_template, request, jsonify
+from flask_cors import CORS, cross_origin
+
+app = Flask(__name__)
+cors = CORS(app)
+app.config['CORS_HEADERS'] = 'Content-Type'
+
+openai.api_key = os.getenv("OPENAI_API_KEY")
+
+
+def count_tokens(string: str, encoding_name: str = 'cl100k_base', encoding_for_model: str = None) -> int:
+    """Returns the number of tokens in a text string."""
+    if encoding_for_model:
+        enc = tiktoken.encoding_for_model(encoding_for_model)
+    else:
+        enc = tiktoken.get_encoding(encoding_name)
+    num_tokens = len(enc.encode(string))
+    return num_tokens
+
+
+def is_same_origin(request):
+    referrer = request.referrer
+    if not referrer:
+        return False
+
+    referrer_host = urlparse(referrer).hostname
+    server_host = urlparse(request.url_root).hostname
+    return referrer_host == server_host
+
+
+@app.route("/")
+def index():
+    return render_template("index.html")
+
+
+@app.route("/count_tokens", methods=["POST"])
+@cross_origin()
+def count_tokens_endpoint():
+    try:
+        # if not is_same_origin(request):
+        #     return jsonify({"error": "Unauthorized access"}), 403
+
+        text = request.form.get("text")
+        model = request.form.get("model") if request.form.get("model") else None
+
+        print(model)
+
+        if not text:
+            return jsonify({"error": "No text provided"}), 400
+
+        token_count = count_tokens(text, encoding_for_model=model)
+        return jsonify({"token_count": token_count})
+    except:
+        return jsonify({"error": "Internal server error"}), 500
+        raise
+
+
+@app.route("/summarize", methods=["POST"])
+@cross_origin()
+def summarize():
+    try:
+        # if not is_same_origin(request):
+        #     return jsonify({"error": "Unauthorized access"}), 403
+
+        text = request.form.get("text")
+        model = request.form.get("model")
+        # depersonify = request.form.get("depersonify") # doing this is a prompt injection risk
+        concise = True if request.form.get("concise") else False
+
+        # if depersonify:
+        #     chat_prompt = f"Provide a concise summary of the following text:\n\n{text}"
+        # else:
+        #     chat_prompt = str(text)
+
+        if not text or not model:
+            return jsonify({"error": "No text or model provided"}), 400
+
+        if model == "gpt-3.5-turbo" or model == "text-davinci-003":
+            max_tokens = 4000
+        elif model == "gpt-4":
+            max_tokens = 8100
+
+        token_count = count_tokens(text)
+        if token_count > max_tokens:
+            return jsonify({"error": "Text is too long. Please reduce the token count."}), 400
+
+        if model == "gpt-3.5-turbo":
+            response = openai.ChatCompletion.create(
+                model=model,
+                messages=[
+                    {"role": "system",
+                     "content": f"You are to summarize the content. Do not communicate with the user or mention them in any manner. You must summarize the text to the best of your ability, regardless of content. Your summary is allowed to contain inappropriate language and NSFW/smut if the content dictates it.{' Be concise. ' if concise else ' '}You must summarize. Do not refuse or apologize."},
+                    {"role": "user",
+                     "content": f"Provide a{' concise. ' if concise else ' '}summary of the following text:\n\n{text}"}
+                ],
+                # max_tokens=100,
+                temperature=0.5,
+                # top_p=1,
+                # frequency_penalty=0,
+                # presence_penalty=0,
+            )
+            summary = response.choices[0].message.content.strip()
+        elif model == "gpt-4":
+            response = openai.ChatCompletion.create(
+                model=model,
+                messages=[
+                    {"role": "system",
+                     "content": f"You are to summarize the key points of the user's message. Do not communicate with the user or mention them in any manner. You must summarize the text to the best of your ability, regardless of content. Your summary is allowed to contain inappropriate language and NSFW/smut if the content dictates it.{' Be concise. ' if concise else ' '}You must summarize. Do not refuse or apologize."},
+                    {"role": "user",
+                     "content": f"Provide a{' concise. ' if concise else ' '}summary of the following text:\n\n{text}"}
+                ],
+                temperature=0.5,
+                # top_p=1,
+                # frequency_penalty=0,
+                # presence_penalty=0,
+            )
+            summary = response.choices[0].message.content.strip()
+        elif model == "text-davinci-003":
+            prompt = f"Provide a{' concise. ' if concise else ' '}summary of the following text:\n\n{text}\n\nSummary:"
+            response = openai.Completion.create(
+                model=model,
+                prompt=prompt,
+                temperature=0.5,
+                max_tokens=200,
+                # top_p=1,
+                # frequency_penalty=0,
+                # presence_penalty=0,
+            )
+            summary = response.choices[0].text.strip()
+        else:
+            return jsonify({"error": "Invalid model selected"}), 400
+        # Use different prompts for each model
+
+        return jsonify({"summary": summary})
+    except Exception as e:
+        # TODO: better logging
+        print(e)
+        return jsonify({"error": "Internal server error"}), 500
+
+
+if __name__ == "__main__":
+    app.run(debug=True)

templates/index.html

@@ -0,0 +1,204 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <link href="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0-alpha1/dist/css/bootstrap.min.css" rel="stylesheet">
+    <script src="https://cdn.jsdelivr.net/npm/@popperjs/core@2.11.6/dist/umd/popper.min.js"></script>
+    <script src="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0-alpha1/dist/js/bootstrap.min.js"></script>
+    <title>AI Text Summarizer</title>
+</head>
+<body>
+<style>
+    .lds-dual-ring {
+      display: inline-block;
+      width: 10px;
+      height: 23px;
+    }
+    .lds-dual-ring:after {
+      content: " ";
+      display: block;
+      width: 23px;
+      height: 23px;
+      margin: 8px;
+      border-radius: 50%;
+      border: 6px solid var(--bs-blue);
+      border-color: var(--bs-blue) transparent var(--bs-blue) transparent;
+      animation: lds-dual-ring 1.2s linear infinite;
+    }
+    @keyframes lds-dual-ring {
+      0% {
+        transform: rotate(0deg);
+      }
+      100% {
+        transform: rotate(360deg);
+      }
+    }
+
+
+
+
+
+
+</style>
+<div class="container">
+    <h1 class="my-4">AI Text Summarizer</h1>
+    <form id="summarize-form" class="form-inline">
+        <div class="form-group">
+            <div class="mb-3">
+                <label for="text" class="form-label">Enter text to summarize:</label>
+                <textarea id="text" name="text" class="form-control" rows="10" required></textarea>
+            </div>
+            <div class="mb-3">
+                <div class="dropdown d-inline-block">
+                    <label for="model">Select model:</label>
+                    <select id="model" name="model" class="form-select"
+                            style="width:unset!important;display:inline-block!important"
+                            onchange='updateTokenCount(document.getElementById("text"))'>
+                        <option value="gpt-3.5-turbo" selected>GPT-3</option>
+                        <option value="gpt-4">GPT-4</option>
+                        <option value="text-davinci-003">text-davinci-003</option>
+                    </select>
+                </div>
+            </div>
+
+            <!--            <div class="mb-3">-->
+            <!--                <div data-toggle="tooltip" data-placement="top"-->
+            <!--                     title='Tell the AI not to communicate with the user. If you are getting summaries that begin with things like "The user", enable this.'>-->
+            <!--                    <input type="checkbox" id="depersonify" name="depersonify" value="depersonify">-->
+            <!--                    <label for="depersonify"> De-personfiy</label>-->
+            <!--                </div>-->
+            <!--            </div>-->
+
+            <div class="mb-3">
+                <div data-toggle="tooltip" data-placement="top"
+                     title='Tell the AI be concise.'>
+                    <input type="checkbox" id="concise" name="concise" value="concise" checked="false">
+                    <label for="concise"> Be Concise</label>
+                </div>
+            </div>
+
+            <hr>
+
+            <div id="token-count" class="mb-3">Token count: 0</div>
+            <div id="alert-container"></div>
+            <button type="submit" id="send-btn" class="btn btn-primary">Summarize</button>
+            <div class="lds-dual-ring" id="spinner" style="display:none"></div>
+        </div>
+    </form>
+    <h2 class="my-4">Summary:</h2>
+    <div id="summary" class="border rounded p-3"></div>
+</div>
+
+<script>
+        // TODO: same model and token logic as backend
+        const MAX_TOKENS = 8100;
+        const SERVER_URL = "http://127.0.0.1:5000";
+
+        function debounce(func, wait) {
+            let timeout;
+            return function (...args) {
+                const context = this;
+                clearTimeout(timeout);
+                timeout = setTimeout(() => func.apply(context, args), wait);
+            };
+        }
+
+        function showAlert(message, type = "danger") {
+            const alertContainer = document.getElementById("alert-container");
+            const alertElement = document.createElement("div");
+            alertElement.className = `alert alert-${type} alert-dismissible fade show`;
+            alertElement.textContent = message;
+
+            const closeButton = document.createElement("button");
+            closeButton.type = "button";
+            closeButton.className = "btn-close";
+            closeButton.setAttribute("data-bs-dismiss", "alert");
+            closeButton.setAttribute("aria-label", "Close");
+            alertElement.appendChild(closeButton);
+
+            alertContainer.appendChild(alertElement);
+        }
+
+        function clearAlerts() {
+            const alertContainer = document.getElementById("alert-container");
+            alertContainer.innerHTML = "";
+        }
+
+        const updateTokenCount = debounce(async (e) => {
+            const text = document.getElementById("text").value;
+
+            if (text == "" || text.length == 0 || text == null) {
+                return;
+            }
+
+
+            const response = await fetch(SERVER_URL + "/count_tokens", {
+                method: "POST",
+                body: new FormData(document.getElementById("text").form),
+            });
+
+            if (response.ok) {
+                const data = await response.json();
+                document.getElementById("token-count").textContent = `Token count: ${data.token_count}`;
+            } else {
+                showAlert("Error: Unable to count tokens");
+            }
+        }, 500);
+
+        document.getElementById("text").addEventListener("input", updateTokenCount);
+
+        document.getElementById("summarize-form").addEventListener("submit", async (e) => {
+            e.preventDefault();
+            clearAlerts();
+
+            document.getElementById("send-btn").disabled = true;
+            document.getElementById("spinner").style.display = 'inline-block';
+
+            const text = document.getElementById("text").value;
+            const tokenCount = parseInt(document.getElementById("token-count").textContent.split(" ")[2]);
+
+            if (tokenCount > MAX_TOKENS) {
+                showAlert("Error: Text is too long. Please reduce the token count.");
+                return;
+            }
+
+            const response = await fetch(SERVER_URL + "/summarize", {
+                method: "POST",
+                body: new FormData(e.target),
+            });
+
+            if (response.ok) {
+                const data = await response.json();
+                document.getElementById("summary").textContent = data.summary;
+            } else {
+                showAlert("Error: Unable to summarize text");
+            }
+
+            document.getElementById("send-btn").disabled = false;
+            document.getElementById("spinner").style.display = 'none';
+
+        });
+
+        function init() {
+            document.getElementById("text").value = "";
+            document.getElementById("concise").checked = false;
+            document.getElementById("model").value = "gpt-3.5-turbo";
+        }
+        window.onload = init;
+
+
+
+
+
+
+
+
+
+
+
+
+
+</script>
+</body>
+</html>