Upload 3 files

Dockerfile

@@ -0,0 +1,12 @@
+FROM python:3.11-slim
+
+WORKDIR /app
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY . .
+
+# Hugging Face exposes port 7860
+EXPOSE 7860
+
+CMD ["python", "main.py"]

main.py

@@ -0,0 +1,196 @@
+import sys
+import os
+import json
+import time
+import requests
+import asyncio
+import threading
+from fastapi import FastAPI, Request, HTTPException, Depends
+from fastapi.responses import JSONResponse
+from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
+import uvicorn
+from pydantic import BaseModel
+from typing import List, Optional
+
+try:
+    from vietnamese_ai.serving.batch_server import MayChuBatch  # type: ignore
+    from vietnamese_ai.llm.vietnamese_llm import VietnameseLLM  # type: ignore
+except ImportError as e:
+    print(f"Error importing vietnamese-ai: {e}. Please run 'pip install vietnamese-ai'")
+    sys.exit(1)
+
+app = FastAPI(
+    title="EvoNet Data Center - B2B API Gateway",
+    description="OpenAI-compatible Gateway routing traffic to DePIN or Local Extreme Batching",
+    version="2.0.0"
+)
+
+CENTER_URL = os.environ.get('CENTER_URL', 'https://evonet-ai.onrender.com')
+# B2B_SECRET is now just a fallback if backend validation is unavailable
+B2B_SECRET = os.environ.get('B2B_SECRET', 'evonet-b2b-partner-secret')
+
+security = HTTPBearer()
+
+print("[Gateway] Khởi tạo V-Neural Extreme JIT Engine...")
+llm = VietnameseLLM(bac=3)
+llm.huan_luyen(["EvoNet AI là mạng DePIN hàng đầu", "Continuous Batching giúp tăng tốc độ inference", "Mô hình ngôn ngữ lớn hoạt động cực nhanh"], so_vong=1)
+
+def fallback_infer(prompt):
+    return llm.sinh_van_ban(prompt, do_dai=20)
+
+def batch_predict(prompts):
+    """Wrapper cho Continuous Batching xử lý nhiều request cùng lúc"""
+    results = []
+    for p in prompts:
+        res = fallback_infer(p)
+        results.append(f"[Extreme JIT] {res}")
+    return results
+
+batch_server = MayChuBatch(
+    mo_hinh=None,
+    ham_du_doan=batch_predict,
+    kich_thuoc_batch=32,
+    timeout_batch=0.05,
+    so_worker=4
+)
+
+def start_heartbeat(port):
+    while True:
+        try:
+            requests.post(f"{CENTER_URL}/api/v1/admin/datacenter/register", json={
+                "role": "Gateway",
+                "port": port
+            }, timeout=5)
+        except Exception:
+            pass
+        time.sleep(10)
+
+@app.on_event("startup")
+async def startup_event():
+    print("[Gateway] Bắt đầu Server Continuous Batching...")
+    batch_server.bat_dau()
+    port = int(os.environ.get("PORT", 7860))
+    threading.Thread(target=start_heartbeat, args=(port,), daemon=True).start()
+
+@app.on_event("shutdown")
+async def shutdown_event():
+    print("[Gateway] Tắt Server Continuous Batching...")
+    batch_server.dung()
+
+# --- Schemas ---
+class ChatMessage(BaseModel):
+    role: str
+    content: str
+
+class ChatCompletionRequest(BaseModel):
+    model: str = "evonet-extreme-7b"
+    messages: List[ChatMessage]
+    temperature: Optional[float] = 0.7
+    max_tokens: Optional[int] = 512
+
+# --- Dependency: Validate API Key via Center Backend ---
+async def verify_b2b_token(credentials: HTTPAuthorizationCredentials = Depends(security)):
+    token = credentials.credentials
+    # Fast path fallback for demo
+    if token == B2B_SECRET:
+        return token
+    
+    # Validation against EvoNet Backend (Admin or API Key check)
+    try:
+        # Check if the API key is valid by asking the Center Backend
+        res = requests.get(f"{CENTER_URL}/api/v1/auth/validate-key", headers={"Authorization": f"Bearer {token}"}, timeout=2)
+        if res.status_code == 200:
+            return token
+    except Exception as e:
+        print(f"[Gateway] Validate Key Error: {e}")
+        pass
+    
+    raise HTTPException(status_code=401, detail="Invalid API Key or Center Backend unavailable")
+
+@app.get("/")
+def read_root():
+    return {
+        "status": "online", 
+        "service": "EvoNet B2B API Gateway",
+        "message": "Gateway is running. Send POST requests to /v1/chat/completions"
+    }
+
+@app.post("/v1/chat/completions")
+async def chat_completions(req: ChatCompletionRequest, request: Request, token: str = Depends(verify_b2b_token)):
+    if not req.messages:
+        raise HTTPException(status_code=400, detail="messages array cannot be empty")
+        
+    last_message = req.messages[-1].content
+    print(f"[Gateway] B2B Request for model {req.model}")
+    
+    active_nodes = 0
+    try:
+        # Step 1: Check swarm health
+        res = requests.get(f"{CENTER_URL}/api/v1/workers", timeout=2)
+        if res.status_code == 200:
+            workers = res.json().get('workers', [])
+            active_nodes = len(workers)
+    except Exception as e:
+        print(f"[Gateway] Center connection failed: {e}")
+        
+    result_text = ""
+    
+    if active_nodes > 0:
+        # Step 2a: Route to DePIN Swarm
+        print(f"[Gateway Routing] ➡️ Chuyển hướng tới DePIN Swarm ({active_nodes} nodes đang rảnh)")
+        try:
+            # We use a loop asyncio to not block FastAPI
+            loop = asyncio.get_event_loop()
+            def call_swarm():
+                return requests.post(
+                    f"{CENTER_URL}/api/v1/swarm-inference", 
+                    json={"prompt": last_message, "max_tokens": req.max_tokens}, 
+                    headers={"Authorization": f"Bearer {token}"}, 
+                    timeout=15
+                )
+            res = await loop.run_in_executor(None, call_swarm)
+            
+            if res.status_code == 200:
+                result_text = res.json().get('result', '')
+            else:
+                raise Exception("Swarm returned non-200")
+        except Exception as e:
+            print(f"[Gateway Routing] Lỗi từ DePIN Swarm ({e}), chuyển sang Fallback Local.")
+            result_text = batch_server.du_doan(last_message)
+    else:
+        # Step 2b: Fallback to Local Extreme Batching
+        print(f"[Gateway Routing] ➡️ DePIN bận/thiếu Node. Chuyển hướng xử lý tại Cụm Server Nội bộ.")
+        result_text = batch_server.du_doan(last_message)
+
+    # Step 3: Format OpenAI response
+    prompt_tokens = len(last_message.split())
+    completion_tokens = len(result_text.split())
+    
+    return {
+        "id": f"chatcmpl-{int(time.time()*1000)}",
+        "object": "chat.completion",
+        "created": int(time.time()),
+        "model": req.model,
+        "choices": [{
+            "index": 0,
+            "message": {
+                "role": "assistant",
+                "content": result_text
+            },
+            "finish_reason": "stop"
+        }],
+        "usage": {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": prompt_tokens + completion_tokens
+        }
+    }
+
+if __name__ == "__main__":
+    port = int(os.environ.get("PORT", 7860))
+    print("=================================================")
+    print("🏢 EvoNet Data Center - B2B API Gateway (FastAPI)")
+    print("⚡ V-Neural Extreme Continuous Batching ACTIVE")
+    print(f"✅ Running on port {port}")
+    print("=================================================")
+    uvicorn.run("main:app", host="0.0.0.0", port=port, log_level="info", workers=1)

requirements.txt

@@ -0,0 +1,7 @@
+fastapi
+uvicorn
+pydantic
+requests
+numpy
+torch
+vietnamese-ai