Spaces:
Sleeping
Sleeping
Create app.py
Browse filesAdded main app.py
app.py
ADDED
|
@@ -0,0 +1,96 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import requests
|
| 3 |
+
from flask import Flask, request, jsonify
|
| 4 |
+
from llama_cpp import Llama
|
| 5 |
+
import subprocess
|
| 6 |
+
import time
|
| 7 |
+
import json
|
| 8 |
+
|
| 9 |
+
app = Flask(__name__)
|
| 10 |
+
MODEL_PATH = "model/calme-3.3-llamaloi-3b.Q4_K_M.gguf"
|
| 11 |
+
GH_PAT = os.getenv("GH_PAT","ghp_TGVMURDWG3HkLIfWUa4hyJUGYOYgXH1EljIi") # GitHub Personal Access Token
|
| 12 |
+
REPO_URL = "https://github.com/NitinBot001/Audio-url-new-js.git"
|
| 13 |
+
|
| 14 |
+
def download_model():
|
| 15 |
+
os.makedirs("model", exist_ok=True)
|
| 16 |
+
if not os.path.exists(MODEL_PATH):
|
| 17 |
+
print("Downloading model...")
|
| 18 |
+
r = requests.get(
|
| 19 |
+
"https://huggingface.co/MaziyarPanahi/calme-3.3-llamaloi-3b-GGUF/resolve/main/calme-3.3-llamaloi-3b.Q4_K_M.gguf",
|
| 20 |
+
stream=True,
|
| 21 |
+
)
|
| 22 |
+
with open(MODEL_PATH, "wb") as f:
|
| 23 |
+
for chunk in r.iter_content(chunk_size=8192):
|
| 24 |
+
f.write(chunk)
|
| 25 |
+
|
| 26 |
+
def start_tunnel():
|
| 27 |
+
# Start nport tunnel
|
| 28 |
+
tunnel_process = subprocess.Popen(
|
| 29 |
+
["npx", "nport", "-s", "ai-service", "-p", "5000"],
|
| 30 |
+
stdout=subprocess.PIPE,
|
| 31 |
+
stderr=subprocess.PIPE,
|
| 32 |
+
)
|
| 33 |
+
time.sleep(10) # Wait for tunnel to establish
|
| 34 |
+
|
| 35 |
+
# Extract tunnel URL from logs
|
| 36 |
+
tunnel_url = None
|
| 37 |
+
for line in iter(tunnel_process.stdout.readline, b""):
|
| 38 |
+
line = line.decode("utf-8").strip()
|
| 39 |
+
if "your domain is:" in line:
|
| 40 |
+
tunnel_url = line.split("your domain is: ")[1]
|
| 41 |
+
break
|
| 42 |
+
|
| 43 |
+
if not tunnel_url:
|
| 44 |
+
raise Exception("Failed to extract tunnel URL")
|
| 45 |
+
|
| 46 |
+
return tunnel_url
|
| 47 |
+
|
| 48 |
+
def push_tunnel_url_to_repo(tunnel_url):
|
| 49 |
+
# Create instance.json
|
| 50 |
+
instance_data = {"tunnel_url": tunnel_url}
|
| 51 |
+
with open("instance.json", "w") as f:
|
| 52 |
+
json.dump(instance_data, f)
|
| 53 |
+
|
| 54 |
+
# Clone the repository
|
| 55 |
+
subprocess.run(
|
| 56 |
+
["git", "clone", f"https://x-access-token:{GH_PAT}@{REPO_URL.split('https://')[1]}", "repo"],
|
| 57 |
+
check=True,
|
| 58 |
+
)
|
| 59 |
+
os.chdir("repo")
|
| 60 |
+
|
| 61 |
+
# Move instance.json to the repository
|
| 62 |
+
subprocess.run(["mv", "../instance.json", "."], check=True)
|
| 63 |
+
|
| 64 |
+
# Commit and push changes
|
| 65 |
+
subprocess.run(["git", "config", "--global", "user.email", "[email protected]"], check=True)
|
| 66 |
+
subprocess.run(["git", "config", "--global", "user.name", "github-actions"], check=True)
|
| 67 |
+
subprocess.run(["git", "add", "instance.json"], check=True)
|
| 68 |
+
subprocess.run(["git", "commit", "-m", f"Update tunnel URL to {tunnel_url}"], check=True)
|
| 69 |
+
subprocess.run(["git", "push", "origin", "main"], check=True)
|
| 70 |
+
|
| 71 |
+
@app.route("/chat", methods=["POST"])
|
| 72 |
+
def chat():
|
| 73 |
+
data = request.json
|
| 74 |
+
prompt = f"<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n{data.get('message','')}<|eot_id|>\n<|start_header_id|>assistant<|end_header_id|>\n"
|
| 75 |
+
output = llm(prompt, max_tokens=2048, stop=["<|eot_id|>"], temperature=0.8, top_p=0.9)
|
| 76 |
+
return jsonify({"response": output["choices"][0]["text"].strip()})
|
| 77 |
+
|
| 78 |
+
if __name__ == "__main__":
|
| 79 |
+
# Download the model
|
| 80 |
+
download_model()
|
| 81 |
+
|
| 82 |
+
# Initialize the LLM
|
| 83 |
+
llm = Llama(
|
| 84 |
+
model_path=MODEL_PATH,
|
| 85 |
+
n_ctx=8192,
|
| 86 |
+
n_threads=2,
|
| 87 |
+
n_gpu_layers=0,
|
| 88 |
+
verbose=False,
|
| 89 |
+
)
|
| 90 |
+
|
| 91 |
+
# Start the tunnel and push the URL
|
| 92 |
+
tunnel_url = start_tunnel()
|
| 93 |
+
push_tunnel_url_to_repo(tunnel_url)
|
| 94 |
+
|
| 95 |
+
# Run the Flask app
|
| 96 |
+
app.run(host="0.0.0.0", port=5000)
|