Spaces:
Running
Running
fix7
Browse files- .dockerignore +8 -0
- Dockerfile +16 -17
- app.py +48 -44
- docker-compose.yml +13 -33
.dockerignore
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
.env
|
2 |
+
*.pyc
|
3 |
+
__pycache__/
|
4 |
+
data/
|
5 |
+
.cache/
|
6 |
+
.vscode/
|
7 |
+
*.log
|
8 |
+
tmp/
|
Dockerfile
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
-
# صورة بايثون خفيفة
|
2 |
FROM python:3.10-slim
|
3 |
|
4 |
# تثبيت الأدوات الأساسية
|
@@ -7,35 +7,34 @@ RUN apt-get update && apt-get install -y \
|
|
7 |
cmake \
|
8 |
gcc \
|
9 |
g++ \
|
|
|
10 |
&& rm -rf /var/lib/apt/lists/*
|
11 |
|
12 |
-
# إنشاء مستخدم غير root
|
13 |
RUN useradd -m -u 1000 user
|
14 |
|
15 |
-
# مجلد العمل
|
16 |
WORKDIR /home/user/app
|
17 |
-
|
18 |
-
# نسخ الملفات إلى الحاوية وتغيير الملكية
|
19 |
COPY --chown=user . .
|
20 |
|
21 |
-
# إنشاء
|
22 |
-
RUN mkdir -p /home/user/app/data/cache &&
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
|
27 |
-
#
|
28 |
-
RUN chown -R user:user /tmp
|
29 |
-
|
30 |
-
# إضافة هذه الخطوة قبل تثبيت المتطلبات
|
31 |
RUN python -m venv /home/user/venv
|
32 |
ENV PATH="/home/user/venv/bin:$PATH"
|
33 |
|
34 |
-
#
|
35 |
RUN pip install --upgrade pip && \
|
36 |
pip install -r requirements.txt && \
|
37 |
echo "✅ تثبيت المتطلبات ناجح" > /tmp/requirements_install.log || \
|
38 |
echo "❌ فشل تثبيت المتطلبات" > /tmp/requirements_install.log
|
39 |
|
40 |
-
#
|
41 |
-
|
|
|
|
|
|
|
|
1 |
+
# استخدام صورة بايثون خفيفة
|
2 |
FROM python:3.10-slim
|
3 |
|
4 |
# تثبيت الأدوات الأساسية
|
|
|
7 |
cmake \
|
8 |
gcc \
|
9 |
g++ \
|
10 |
+
python3-dev \
|
11 |
&& rm -rf /var/lib/apt/lists/*
|
12 |
|
13 |
+
# إنشاء مستخدم غير root
|
14 |
RUN useradd -m -u 1000 user
|
15 |
|
16 |
+
# إنشاء مجلد العمل وتعيين الصلاحيات
|
17 |
WORKDIR /home/user/app
|
|
|
|
|
18 |
COPY --chown=user . .
|
19 |
|
20 |
+
# إنشاء المجلدات المطلوبة وتعيين الصلاحيات
|
21 |
+
RUN mkdir -p /home/user/app/data/cache && \
|
22 |
+
mkdir -p /home/user/app/data && \
|
23 |
+
chown -R user:user /home/user/app/data && \
|
24 |
+
chown -R user:user /tmp
|
25 |
|
26 |
+
# تفعيل بيئة افتراضية في home وليس داخل المشروع
|
|
|
|
|
|
|
27 |
RUN python -m venv /home/user/venv
|
28 |
ENV PATH="/home/user/venv/bin:$PATH"
|
29 |
|
30 |
+
# تثبيت المتطلبات
|
31 |
RUN pip install --upgrade pip && \
|
32 |
pip install -r requirements.txt && \
|
33 |
echo "✅ تثبيت المتطلبات ناجح" > /tmp/requirements_install.log || \
|
34 |
echo "❌ فشل تثبيت المتطلبات" > /tmp/requirements_install.log
|
35 |
|
36 |
+
# استخدام المستخدم غير root
|
37 |
+
USER user
|
38 |
+
|
39 |
+
# تشغيل التطبيق
|
40 |
+
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]
|
app.py
CHANGED
@@ -1,15 +1,18 @@
|
|
1 |
-
from fastapi import FastAPI
|
2 |
from pydantic import BaseModel
|
3 |
from llama_cpp import Llama
|
4 |
import logging
|
5 |
import os
|
6 |
import threading
|
7 |
from fastapi.middleware.cors import CORSMiddleware
|
8 |
-
|
9 |
-
# استيراد وحدة المراقبة المعدلة
|
10 |
from monitor import get_current_metrics, start_monitoring_thread
|
|
|
|
|
|
|
|
|
|
|
11 |
|
12 |
-
#
|
13 |
logging.basicConfig(
|
14 |
level=logging.DEBUG,
|
15 |
format="🪵 [%(asctime)s] [%(levelname)s] %(message)s",
|
@@ -19,60 +22,65 @@ logger = logging.getLogger(__name__)
|
|
19 |
MODEL_REPO = "QuantFactory/Qwen2.5-7B-Instruct-GGUF"
|
20 |
MODEL_FILE = "Qwen2.5-7B-Instruct.Q4_K_M.gguf"
|
21 |
MODEL_PATH = f"/home/user/app/data/cache/{MODEL_FILE}"
|
|
|
22 |
|
|
|
23 |
if not os.path.exists(MODEL_PATH):
|
24 |
-
token=os.getenv("HF_TOKEN")
|
25 |
-
from huggingface_hub import hf_hub_download
|
26 |
os.makedirs("/home/user/app/data/cache", exist_ok=True)
|
27 |
logger.info("📦 تحميل النموذج من Hugging Face Hub...")
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
33 |
|
|
|
34 |
if os.path.exists(MODEL_PATH):
|
35 |
logger.info(f"✅ النموذج موجود: {MODEL_PATH}")
|
36 |
else:
|
37 |
logger.error(f"❌ النموذج غير موجود: {MODEL_PATH}")
|
|
|
38 |
|
39 |
# تحميل النموذج
|
40 |
llm = Llama(
|
41 |
model_path=MODEL_PATH,
|
42 |
-
n_ctx=2048,
|
43 |
-
n_threads=4,
|
44 |
n_gpu_layers=0,
|
45 |
-
n_batch=512,
|
46 |
-
use_mlock=True,
|
47 |
verbose=False
|
48 |
)
|
49 |
|
50 |
-
#
|
51 |
try:
|
52 |
-
|
53 |
-
logger.info("🔍 يجري اختبار النموذج...")
|
54 |
test_output = llm("مرحبا", max_tokens=10)
|
55 |
logger.info(f"✅ اختبار النموذج ناجح: {test_output}")
|
56 |
except Exception as e:
|
57 |
logger.error(f"❌ فشل اختبار النموذج: {str(e)}")
|
58 |
-
raise RuntimeError("فشل
|
59 |
|
60 |
SYSTEM_PROMPT = """<|im_start|>system
|
61 |
You are Qwen, created by Alibaba Cloud. You are an AI development assistant. Follow these rules:
|
62 |
1. If request is simple (single file, <50 lines), handle it directly
|
63 |
-
2. For complex requests (multiple files, >50 lines), just respond with
|
64 |
3. Always check code for errors before sending
|
65 |
4. Never execute unsafe code<|im_end|>"""
|
66 |
|
67 |
-
# بدء
|
68 |
start_monitoring_thread()
|
69 |
|
70 |
-
# API setup
|
71 |
app = FastAPI()
|
72 |
|
73 |
app.add_middleware(
|
74 |
CORSMiddleware,
|
75 |
-
allow_origins=["*"], #
|
76 |
allow_credentials=True,
|
77 |
allow_methods=["*"],
|
78 |
allow_headers=["*"],
|
@@ -84,11 +92,11 @@ async def startup_event():
|
|
84 |
|
85 |
class ChatRequest(BaseModel):
|
86 |
message: str
|
87 |
-
history: list[
|
88 |
|
89 |
class ChatResponse(BaseModel):
|
90 |
response: str
|
91 |
-
updated_history: list[
|
92 |
|
93 |
def format_prompt(messages):
|
94 |
formatted = []
|
@@ -99,7 +107,7 @@ def format_prompt(messages):
|
|
99 |
formatted.append(f"<|im_start|>user\n{content}<|im_end|>")
|
100 |
else:
|
101 |
formatted.append(f"<|im_start|>assistant\n{content}<|im_end|>")
|
102 |
-
formatted.append("<|im_start|>assistant\n")
|
103 |
return "\n".join(formatted)
|
104 |
|
105 |
@app.get("/metrics")
|
@@ -110,20 +118,18 @@ def read_metrics():
|
|
110 |
@app.post("/chat", response_model=ChatResponse)
|
111 |
def chat(req: ChatRequest):
|
112 |
logger.info(f"📩 طلب جديد: {req.message}")
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
messages.append(("
|
119 |
-
messages.append(("user", req.message))
|
120 |
|
121 |
-
|
122 |
-
|
123 |
|
124 |
-
try:
|
125 |
import gc
|
126 |
-
gc.collect()
|
127 |
output = llm(
|
128 |
prompt,
|
129 |
max_tokens=1024,
|
@@ -134,14 +140,12 @@ def chat(req: ChatRequest):
|
|
134 |
)
|
135 |
reply = output["choices"][0]["text"].strip()
|
136 |
logger.info(f"🤖 رد النموذج: {reply}")
|
|
|
|
|
137 |
except Exception as e:
|
138 |
-
logger.error(f"
|
139 |
-
raise
|
140 |
-
|
141 |
-
# إصلاح تحديث السجل
|
142 |
-
updated_history = req.history + [(req.message, reply)]
|
143 |
-
return ChatResponse(response=reply, updated_history=updated_history)
|
144 |
|
145 |
@app.get("/")
|
146 |
def root():
|
147 |
-
return {"message": "الخادم يعمل", "status": "ok"}
|
|
|
1 |
+
from fastapi import FastAPI, HTTPException
|
2 |
from pydantic import BaseModel
|
3 |
from llama_cpp import Llama
|
4 |
import logging
|
5 |
import os
|
6 |
import threading
|
7 |
from fastapi.middleware.cors import CORSMiddleware
|
|
|
|
|
8 |
from monitor import get_current_metrics, start_monitoring_thread
|
9 |
+
from huggingface_hub import hf_hub_download
|
10 |
+
from dotenv import load_dotenv
|
11 |
+
|
12 |
+
# تحميل متغيرات البيئة
|
13 |
+
load_dotenv()
|
14 |
|
15 |
+
# إعداد السجل
|
16 |
logging.basicConfig(
|
17 |
level=logging.DEBUG,
|
18 |
format="🪵 [%(asctime)s] [%(levelname)s] %(message)s",
|
|
|
22 |
MODEL_REPO = "QuantFactory/Qwen2.5-7B-Instruct-GGUF"
|
23 |
MODEL_FILE = "Qwen2.5-7B-Instruct.Q4_K_M.gguf"
|
24 |
MODEL_PATH = f"/home/user/app/data/cache/{MODEL_FILE}"
|
25 |
+
HF_TOKEN = os.getenv("HF_TOKEN")
|
26 |
|
27 |
+
# تحميل النموذج إذا لم يكن موجودًا
|
28 |
if not os.path.exists(MODEL_PATH):
|
|
|
|
|
29 |
os.makedirs("/home/user/app/data/cache", exist_ok=True)
|
30 |
logger.info("📦 تحميل النموذج من Hugging Face Hub...")
|
31 |
+
try:
|
32 |
+
hf_hub_download(
|
33 |
+
repo_id=MODEL_REPO,
|
34 |
+
filename=MODEL_FILE,
|
35 |
+
local_dir="/home/user/app/data/cache",
|
36 |
+
token=HF_TOKEN,
|
37 |
+
)
|
38 |
+
except Exception as e:
|
39 |
+
logger.error(f"❌ فشل تحميل النموذج: {str(e)}")
|
40 |
+
raise RuntimeError("فشل تحميل النموذج") from e
|
41 |
|
42 |
+
# تأكيد وجود النموذج
|
43 |
if os.path.exists(MODEL_PATH):
|
44 |
logger.info(f"✅ النموذج موجود: {MODEL_PATH}")
|
45 |
else:
|
46 |
logger.error(f"❌ النموذج غير موجود: {MODEL_PATH}")
|
47 |
+
raise RuntimeError("النموذج غير موجود بعد التحميل")
|
48 |
|
49 |
# تحميل النموذج
|
50 |
llm = Llama(
|
51 |
model_path=MODEL_PATH,
|
52 |
+
n_ctx=2048,
|
53 |
+
n_threads=4,
|
54 |
n_gpu_layers=0,
|
55 |
+
n_batch=512,
|
56 |
+
use_mlock=True,
|
57 |
verbose=False
|
58 |
)
|
59 |
|
60 |
+
# اختبار النموذج مباشرة
|
61 |
try:
|
62 |
+
logger.info("🔍 اختبار النموذج...")
|
|
|
63 |
test_output = llm("مرحبا", max_tokens=10)
|
64 |
logger.info(f"✅ اختبار النموذج ناجح: {test_output}")
|
65 |
except Exception as e:
|
66 |
logger.error(f"❌ فشل اختبار النموذج: {str(e)}")
|
67 |
+
raise RuntimeError("فشل اختبار النموذج") from e
|
68 |
|
69 |
SYSTEM_PROMPT = """<|im_start|>system
|
70 |
You are Qwen, created by Alibaba Cloud. You are an AI development assistant. Follow these rules:
|
71 |
1. If request is simple (single file, <50 lines), handle it directly
|
72 |
+
2. For complex requests (multiple files, >50 lines), just respond with "CODER"
|
73 |
3. Always check code for errors before sending
|
74 |
4. Never execute unsafe code<|im_end|>"""
|
75 |
|
76 |
+
# بدء مراقبة الموارد
|
77 |
start_monitoring_thread()
|
78 |
|
|
|
79 |
app = FastAPI()
|
80 |
|
81 |
app.add_middleware(
|
82 |
CORSMiddleware,
|
83 |
+
allow_origins=["*"], # يمكن تخصيصه لاحقًا
|
84 |
allow_credentials=True,
|
85 |
allow_methods=["*"],
|
86 |
allow_headers=["*"],
|
|
|
92 |
|
93 |
class ChatRequest(BaseModel):
|
94 |
message: str
|
95 |
+
history: list[list[str]] = [] # يجب أن تكون قائمة من القوائم لتمثيل JSON
|
96 |
|
97 |
class ChatResponse(BaseModel):
|
98 |
response: str
|
99 |
+
updated_history: list[list[str]]
|
100 |
|
101 |
def format_prompt(messages):
|
102 |
formatted = []
|
|
|
107 |
formatted.append(f"<|im_start|>user\n{content}<|im_end|>")
|
108 |
else:
|
109 |
formatted.append(f"<|im_start|>assistant\n{content}<|im_end|>")
|
110 |
+
formatted.append("<|im_start|>assistant\n")
|
111 |
return "\n".join(formatted)
|
112 |
|
113 |
@app.get("/metrics")
|
|
|
118 |
@app.post("/chat", response_model=ChatResponse)
|
119 |
def chat(req: ChatRequest):
|
120 |
logger.info(f"📩 طلب جديد: {req.message}")
|
121 |
+
try:
|
122 |
+
messages = [("system", SYSTEM_PROMPT)]
|
123 |
+
for user_msg, bot_msg in req.history:
|
124 |
+
messages.append(("user", user_msg))
|
125 |
+
messages.append(("assistant", bot_msg))
|
126 |
+
messages.append(("user", req.message))
|
|
|
127 |
|
128 |
+
prompt = format_prompt(messages)
|
129 |
+
logger.debug(f"📝 prompt المُرسل:\n{prompt[:300]}...")
|
130 |
|
|
|
131 |
import gc
|
132 |
+
gc.collect()
|
133 |
output = llm(
|
134 |
prompt,
|
135 |
max_tokens=1024,
|
|
|
140 |
)
|
141 |
reply = output["choices"][0]["text"].strip()
|
142 |
logger.info(f"🤖 رد النموذج: {reply}")
|
143 |
+
updated_history = req.history + [[req.message, reply]]
|
144 |
+
return ChatResponse(response=reply, updated_history=updated_history)
|
145 |
except Exception as e:
|
146 |
+
logger.error(f"❌ خطأ أثناء المعالجة: {str(e)}")
|
147 |
+
raise HTTPException(status_code=500, detail="حدث خطأ أثناء توليد الرد")
|
|
|
|
|
|
|
|
|
148 |
|
149 |
@app.get("/")
|
150 |
def root():
|
151 |
+
return {"message": "الخادم يعمل", "status": "ok"}
|
docker-compose.yml
CHANGED
@@ -1,43 +1,23 @@
|
|
1 |
-
version: '3.
|
2 |
|
3 |
services:
|
4 |
ai-assistant:
|
5 |
-
build:
|
|
|
|
|
|
|
6 |
ports:
|
7 |
- "7860:7860"
|
|
|
|
|
8 |
environment:
|
9 |
- HF_TOKEN=${HF_TOKEN}
|
10 |
- GOOGLE_DRIVE_FOLDER_ID=${GOOGLE_DRIVE_FOLDER_ID}
|
11 |
- GITHUB_REPO=${GITHUB_REPO}
|
12 |
- GITHUB_TOKEN=${GITHUB_TOKEN}
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
memory: 8G
|
20 |
-
reservations:
|
21 |
-
memory: 6G
|
22 |
-
command: ["sh", "-c", "echo '✅ تم بدء الخدمة مع استهلاك موارد: $(cat /sys/fs/cgroup/memory/memory.usage_in_bytes)' > /tmp/resource_check.log && uvicorn main:app --host 0.0.0.0 --port 7860"]
|
23 |
-
|
24 |
-
prometheus:
|
25 |
-
image: prom/prometheus
|
26 |
-
volumes:
|
27 |
-
- ./prometheus.yml:/etc/prometheus/prometheus.yml
|
28 |
-
ports:
|
29 |
-
- "9090:9090"
|
30 |
-
command: ["--config.file=/etc/prometheus/prometheus.yml", "--web.enable-lifecycle"]
|
31 |
-
|
32 |
-
grafana:
|
33 |
-
image: grafana/grafana
|
34 |
-
ports:
|
35 |
-
- "3000:3000"
|
36 |
-
environment:
|
37 |
-
- GF_SECURITY_ADMIN_PASSWORD=admin
|
38 |
-
volumes:
|
39 |
-
- grafana-data:/var/lib/grafana
|
40 |
-
command: ["--homepath=/usr/share/grafana"]
|
41 |
-
|
42 |
-
volumes:
|
43 |
-
grafana-data:
|
|
|
1 |
+
version: '3.9'
|
2 |
|
3 |
services:
|
4 |
ai-assistant:
|
5 |
+
build:
|
6 |
+
context: .
|
7 |
+
dockerfile: Dockerfile
|
8 |
+
container_name: ai-dev-assistant
|
9 |
ports:
|
10 |
- "7860:7860"
|
11 |
+
volumes:
|
12 |
+
- ./data:/home/user/app/data # حفظ البيانات والنموذج خارجيًا
|
13 |
environment:
|
14 |
- HF_TOKEN=${HF_TOKEN}
|
15 |
- GOOGLE_DRIVE_FOLDER_ID=${GOOGLE_DRIVE_FOLDER_ID}
|
16 |
- GITHUB_REPO=${GITHUB_REPO}
|
17 |
- GITHUB_TOKEN=${GITHUB_TOKEN}
|
18 |
+
restart: unless-stopped
|
19 |
+
healthcheck:
|
20 |
+
test: ["CMD", "curl", "-f", "http://localhost:7860/"]
|
21 |
+
interval: 30s
|
22 |
+
timeout: 5s
|
23 |
+
retries: 3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|