Update app.py
Browse files
app.py
CHANGED
|
@@ -1,47 +1,90 @@
|
|
| 1 |
-
from fastapi import FastAPI, Request
|
| 2 |
-
from fastapi.responses import StreamingResponse
|
| 3 |
-
from openai import OpenAI
|
| 4 |
import os
|
|
|
|
|
|
|
| 5 |
import chromadb
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
|
| 7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
-
|
| 10 |
-
STORAGE_PATH = "./neural_memory"
|
| 11 |
chroma_client = chromadb.PersistentClient(path=STORAGE_PATH)
|
| 12 |
-
|
|
|
|
| 13 |
|
| 14 |
client = OpenAI(
|
| 15 |
base_url="https://router.huggingface.co/hf-inference/v1",
|
| 16 |
api_key=os.getenv("HF_TOKEN")
|
| 17 |
)
|
| 18 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
@app.post("/v1/chat/completions")
|
| 20 |
-
async def
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
|
|
|
| 24 |
|
| 25 |
-
# البحث
|
|
|
|
| 26 |
results = collection.query(query_texts=[user_query], n_results=3)
|
| 27 |
knowledge = "\n".join(results['documents'][0]) if results['documents'] else ""
|
| 28 |
|
| 29 |
-
# حقن المعرفة
|
| 30 |
-
|
|
|
|
| 31 |
|
| 32 |
-
def
|
| 33 |
response = client.chat.completions.create(
|
| 34 |
model="huihui-ai/Qwen2.5-72B-Instruct-abliterated",
|
| 35 |
messages=messages,
|
|
|
|
| 36 |
stream=True
|
| 37 |
)
|
| 38 |
for chunk in response:
|
| 39 |
-
|
| 40 |
-
|
|
|
|
|
|
|
| 41 |
yield "data: [DONE]\n\n"
|
| 42 |
|
| 43 |
-
return StreamingResponse(
|
| 44 |
|
| 45 |
if __name__ == "__main__":
|
| 46 |
import uvicorn
|
| 47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import os
|
| 2 |
+
import uuid
|
| 3 |
+
import pypdf
|
| 4 |
import chromadb
|
| 5 |
+
from fastapi import FastAPI, Request, File, UploadFile, BackgroundTasks
|
| 6 |
+
from fastapi.responses import StreamingResponse
|
| 7 |
+
from fastapi.middleware.cors import CORSMiddleware
|
| 8 |
+
from openai import OpenAI
|
| 9 |
+
from chromadb.utils import embedding_functions
|
| 10 |
+
|
| 11 |
+
# --- الإعدادات الفنية ---
|
| 12 |
+
app = FastAPI(title="Neural RAG Engine")
|
| 13 |
|
| 14 |
+
# تفعيل CORS للسماح لواجهات Next.js بالاتصال بالخادم
|
| 15 |
+
app.add_middleware(
|
| 16 |
+
CORSMiddleware,
|
| 17 |
+
allow_origins=["*"],
|
| 18 |
+
allow_methods=["*"],
|
| 19 |
+
allow_headers=["*"],
|
| 20 |
+
)
|
| 21 |
|
| 22 |
+
STORAGE_PATH = "/data/neural_memory" if os.path.exists("/data") else "./neural_memory"
|
|
|
|
| 23 |
chroma_client = chromadb.PersistentClient(path=STORAGE_PATH)
|
| 24 |
+
default_ef = embedding_functions.DefaultEmbeddingFunction()
|
| 25 |
+
collection = chroma_client.get_or_create_collection(name="advanced_brain_v6", embedding_function=default_ef)
|
| 26 |
|
| 27 |
client = OpenAI(
|
| 28 |
base_url="https://router.huggingface.co/hf-inference/v1",
|
| 29 |
api_key=os.getenv("HF_TOKEN")
|
| 30 |
)
|
| 31 |
|
| 32 |
+
# --- وظائف المعالجة ---
|
| 33 |
+
def ingest_document(file_content, filename):
|
| 34 |
+
# (نفس منطق التقطيع الاحترافي السابق مع الـ Overlap)
|
| 35 |
+
text = file_content.decode("utf-8", errors="ignore")
|
| 36 |
+
# إذا كان PDF يحتاج لمكتبة pypdf (يمكن دمجها هنا)
|
| 37 |
+
chunk_size, overlap = 1000, 200
|
| 38 |
+
chunks = [text[i : i + chunk_size] for i in range(0, len(text), chunk_size - overlap)]
|
| 39 |
+
ids = [str(uuid.uuid4()) for _ in chunks]
|
| 40 |
+
metadatas = [{"source": filename} for _ in chunks]
|
| 41 |
+
collection.add(documents=chunks, ids=ids, metadatas=metadatas)
|
| 42 |
+
|
| 43 |
+
# --- الـ Endpoints المتوافقة مع Next.js ---
|
| 44 |
+
|
| 45 |
+
@app.get("/")
|
| 46 |
+
def home():
|
| 47 |
+
return {"status": "Neural Engine is Running", "docs_in_memory": collection.count()}
|
| 48 |
+
|
| 49 |
+
@app.post("/upload")
|
| 50 |
+
async def upload_file(background_tasks: BackgroundTasks, file: UploadFile = File(...)):
|
| 51 |
+
content = await file.read()
|
| 52 |
+
background_tasks.add_task(ingest_document, content, file.filename)
|
| 53 |
+
return {"message": f"Processing {file.filename} in background..."}
|
| 54 |
+
|
| 55 |
@app.post("/v1/chat/completions")
|
| 56 |
+
async def chat_endpoint(request: Request):
|
| 57 |
+
"""هذا المسار يجعل الخادم متوافقاً تماماً مع واجهات Next.js"""
|
| 58 |
+
body = await request.json()
|
| 59 |
+
messages = body.get("messages", [])
|
| 60 |
+
temperature = body.get("temperature", 0.7)
|
| 61 |
|
| 62 |
+
# البحث الدلالي (RAG) بناءً على آخر رسالة
|
| 63 |
+
user_query = messages[-1]["content"] if messages else ""
|
| 64 |
results = collection.query(query_texts=[user_query], n_results=3)
|
| 65 |
knowledge = "\n".join(results['documents'][0]) if results['documents'] else ""
|
| 66 |
|
| 67 |
+
# حقن المعرفة
|
| 68 |
+
system_instruction = f"Context from Memory:\n{knowledge}\n\nAnswer based on this context."
|
| 69 |
+
messages.insert(0, {"role": "system", "content": system_instruction})
|
| 70 |
|
| 71 |
+
def stream_gen():
|
| 72 |
response = client.chat.completions.create(
|
| 73 |
model="huihui-ai/Qwen2.5-72B-Instruct-abliterated",
|
| 74 |
messages=messages,
|
| 75 |
+
temperature=temperature,
|
| 76 |
stream=True
|
| 77 |
)
|
| 78 |
for chunk in response:
|
| 79 |
+
content = chunk.choices[0].delta.content
|
| 80 |
+
if content:
|
| 81 |
+
# التنسيق المطلوب لـ Server-Sent Events (SSE)
|
| 82 |
+
yield f"data: {content}\n\n"
|
| 83 |
yield "data: [DONE]\n\n"
|
| 84 |
|
| 85 |
+
return StreamingResponse(stream_gen(), media_type="text/event-stream")
|
| 86 |
|
| 87 |
if __name__ == "__main__":
|
| 88 |
import uvicorn
|
| 89 |
+
# تشغيل الخادم
|
| 90 |
+
uvicorn.run(app, host="0.0.0.0", port=7860)
|