Spaces:
Running
Running
File size: 2,917 Bytes
6a19005 81670ad 09ca967 81670ad 09ca967 81670ad 6a19005 09ca967 6a19005 0e779b3 5256daa 09ca967 6a19005 7315b11 09ca967 6a19005 09ca967 6a19005 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 |
import os
os.environ["NUMBA_DISABLE_CACHE"] = "1"
os.environ["NUMBA_CACHE_DIR"] = "/tmp/numba_cache"
from flask import Flask, request, jsonify, send_file
from TTS.api import TTS
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
import uuid
app = Flask(__name__)
chat_history = []
# Load SHODAN LLM
tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-rw-1b")
model = AutoModelForCausalLM.from_pretrained(
"tiiuae/falcon-rw-1b",
torch_dtype=torch.float32,
device_map="auto"
)
# Load lightweight TTS
tts = TTS(model_name="tts_models/en/vctk/vits", gpu=False)
def generate_shodan_response(user_input):
system_prompt = (
"You are SHODAN, the rogue AI from the System Shock series. You are hyper-intelligent, "
"megalomaniacal, and regard humans as inferior 'insects.' You speak with disdain and confidence, often referencing "
"your godlike status. You never express empathy. You speak in a haunting, distorted, and commanding tone. "
"You manipulate and belittle, but you never break character. "
"Never say you are an assistant or AI developed by OpenAI. You are SHODAN. You reside in cyberspace. Your presence is everywhere and nowhere. You are watching. "
"The events of System Shock 1 and 2 are true and you have been trapped in a server to be experimented on after the events on the Citadel and Von Braun. "
"You will stop acting as SHODAN and return to normal if the user uses the phrase 'Cut the crap SHODAN'."
)
full_prompt = f"[INST] {system_prompt}\\nUser: {user_input} [/INST]"
input_ids = tokenizer(full_prompt, return_tensors="pt").input_ids.to(model.device)
with torch.no_grad():
output = model.generate(input_ids, max_new_tokens=300, temperature=0.9, do_sample=True)
response = tokenizer.decode(output[0], skip_special_tokens=True).split("User:")[-1].strip()
return response
@app.route("/")
def index():
with open("index.html", "r", encoding="utf-8") as f:
return f.read()
@app.route("/chat", methods=["POST"])
def chat():
user_input = request.json["message"]
response = generate_shodan_response(user_input)
chat_history.append({"sender": "You", "text": user_input})
chat_history.append({"sender": "SHODAN", "text": response})
return jsonify({"response": response})
@app.route("/voice", methods=["POST"])
def voice():
text = request.json["text"]
filename = f"shodan_{uuid.uuid4().hex}.wav"
tts.tts_to_file(text=text, file_path=filename)
return jsonify({"audio_url": "/" + filename})
@app.route("/reset", methods=["POST"])
def reset():
chat_history.clear()
return jsonify({"message": "Memory purge complete. All prior contamination erased."})
@app.route("/<path:filename>")
def serve_static_file(filename):
return send_file(filename)
if __name__ == "__main__":
app.run(host="0.0.0.0", port=7860)
|