|
|
from flask import Flask, request, jsonify, render_template_string |
|
|
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, AutoModelForCausalLM |
|
|
import torch |
|
|
from langdetect import detect |
|
|
import re |
|
|
import logging |
|
|
|
|
|
|
|
|
logging.getLogger("werkzeug").setLevel(logging.ERROR) |
|
|
|
|
|
app = Flask(__name__) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
GRAMMAR_MODEL = "vennify/t5-base-grammar-correction" |
|
|
|
|
|
grammar_tokenizer = AutoTokenizer.from_pretrained(GRAMMAR_MODEL) |
|
|
grammar_model = AutoModelForSeq2SeqLM.from_pretrained(GRAMMAR_MODEL) |
|
|
grammar_model.eval() |
|
|
|
|
|
def grammar_check(text): |
|
|
if not text.strip(): |
|
|
return text |
|
|
|
|
|
inputs = grammar_tokenizer( |
|
|
"grammar: " + text, |
|
|
return_tensors="pt", |
|
|
truncation=True |
|
|
) |
|
|
|
|
|
with torch.no_grad(): |
|
|
outputs = grammar_model.generate( |
|
|
**inputs, |
|
|
max_length=256 |
|
|
) |
|
|
|
|
|
return grammar_tokenizer.decode( |
|
|
outputs[0], |
|
|
skip_special_tokens=True |
|
|
) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
CHAT_MODEL = "TinyLlama/TinyLlama-1.1B-Chat-v1.0" |
|
|
|
|
|
chat_tokenizer = AutoTokenizer.from_pretrained(CHAT_MODEL) |
|
|
chat_model = AutoModelForCausalLM.from_pretrained( |
|
|
CHAT_MODEL, |
|
|
dtype=torch.float32 |
|
|
) |
|
|
chat_model.eval() |
|
|
|
|
|
def generate_text(prompt): |
|
|
if not prompt.strip(): |
|
|
return "" |
|
|
|
|
|
inputs = chat_tokenizer( |
|
|
prompt, |
|
|
return_tensors="pt" |
|
|
) |
|
|
|
|
|
with torch.no_grad(): |
|
|
output = chat_model.generate( |
|
|
**inputs, |
|
|
max_new_tokens=150, |
|
|
do_sample=True, |
|
|
temperature=0.7, |
|
|
top_p=0.9 |
|
|
) |
|
|
|
|
|
return chat_tokenizer.decode( |
|
|
output[0], |
|
|
skip_special_tokens=True |
|
|
) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def detect_language(text): |
|
|
try: |
|
|
return detect(text) |
|
|
except: |
|
|
return "unknown" |
|
|
|
|
|
def ai_text_detector(text): |
|
|
words = text.split() |
|
|
if len(words) < 30: |
|
|
return "Not enough text to analyze" |
|
|
|
|
|
avg_len = sum(len(w) for w in words) / len(words) |
|
|
repetition = len(set(words)) / len(words) |
|
|
|
|
|
score = 0 |
|
|
if avg_len > 5.5: |
|
|
score += 1 |
|
|
if repetition < 0.55: |
|
|
score += 1 |
|
|
if re.search( |
|
|
r"\b(in conclusion|overall|moreover|furthermore)\b", |
|
|
text.lower() |
|
|
): |
|
|
score += 1 |
|
|
|
|
|
return "Likely AI-generated" if score >= 2 else "Likely human-written" |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
HTML_PAGE = """ |
|
|
<!DOCTYPE html> |
|
|
<html> |
|
|
<head> |
|
|
<title>Local AI Text Tool</title> |
|
|
<style> |
|
|
body { |
|
|
font-family: Arial, sans-serif; |
|
|
background: #f4f6f8; |
|
|
padding: 20px; |
|
|
} |
|
|
.container { |
|
|
max-width: 900px; |
|
|
margin: auto; |
|
|
background: white; |
|
|
padding: 20px; |
|
|
border-radius: 6px; |
|
|
} |
|
|
textarea { |
|
|
width: 100%; |
|
|
height: 120px; |
|
|
margin-bottom: 10px; |
|
|
padding: 10px; |
|
|
font-size: 14px; |
|
|
} |
|
|
button { |
|
|
padding: 10px 16px; |
|
|
margin-right: 5px; |
|
|
cursor: pointer; |
|
|
} |
|
|
.output { |
|
|
background: #eee; |
|
|
padding: 10px; |
|
|
margin-top: 10px; |
|
|
white-space: pre-wrap; |
|
|
} |
|
|
</style> |
|
|
</head> |
|
|
<body> |
|
|
<div class="container"> |
|
|
<h1>Local AI Text App</h1> |
|
|
|
|
|
<textarea id="text" placeholder="Enter text here..."></textarea> |
|
|
|
|
|
<div> |
|
|
<button onclick="grammar()">Grammar Check</button> |
|
|
<button onclick="language()">Detect Language</button> |
|
|
<button onclick="detectAI()">AI Text Detection</button> |
|
|
<button onclick="generate()">Generate Text</button> |
|
|
</div> |
|
|
|
|
|
<h2>Output</h2> |
|
|
<div class="output" id="output"></div> |
|
|
</div> |
|
|
|
|
|
<script> |
|
|
async function postData(url, data) { |
|
|
const res = await fetch(url, { |
|
|
method: "POST", |
|
|
headers: {"Content-Type": "application/json"}, |
|
|
body: JSON.stringify(data) |
|
|
}); |
|
|
return res.json(); |
|
|
} |
|
|
|
|
|
async function grammar() { |
|
|
const text = document.getElementById("text").value; |
|
|
const res = await postData("/grammar", {text}); |
|
|
document.getElementById("output").innerText = res.corrected_text; |
|
|
} |
|
|
|
|
|
async function language() { |
|
|
const text = document.getElementById("text").value; |
|
|
const res = await postData("/language", {text}); |
|
|
document.getElementById("output").innerText = "Language: " + res.language; |
|
|
} |
|
|
|
|
|
async function detectAI() { |
|
|
const text = document.getElementById("text").value; |
|
|
const res = await postData("/detect-ai", {text}); |
|
|
document.getElementById("output").innerText = res.result; |
|
|
} |
|
|
|
|
|
async function generate() { |
|
|
const text = document.getElementById("text").value; |
|
|
const res = await postData("/generate", {prompt: text}); |
|
|
document.getElementById("output").innerText = res.generated_text; |
|
|
} |
|
|
</script> |
|
|
</body> |
|
|
</html> |
|
|
""" |
|
|
|
|
|
@app.route("/", methods=["GET"]) |
|
|
def home(): |
|
|
return render_template_string(HTML_PAGE) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@app.route("/grammar", methods=["POST"]) |
|
|
def grammar(): |
|
|
return jsonify({ |
|
|
"corrected_text": grammar_check(request.json.get("text", "")) |
|
|
}) |
|
|
|
|
|
@app.route("/language", methods=["POST"]) |
|
|
def language(): |
|
|
return jsonify({ |
|
|
"language": detect_language(request.json.get("text", "")) |
|
|
}) |
|
|
|
|
|
@app.route("/generate", methods=["POST"]) |
|
|
def generate(): |
|
|
return jsonify({ |
|
|
"generated_text": generate_text(request.json.get("prompt", "")) |
|
|
}) |
|
|
|
|
|
@app.route("/detect-ai", methods=["POST"]) |
|
|
def detect_ai(): |
|
|
return jsonify({ |
|
|
"result": ai_text_detector(request.json.get("text", "")) |
|
|
}) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if __name__ == "__main__": |
|
|
app.run(host="0.0.0.0", port=7860) |
|
|
|
