Spaces:
Sleeping
Sleeping
File size: 11,046 Bytes
acd560f | 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 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 | import os
import json
import re
import requests
import threading
import gc
from datetime import datetime
from bs4 import BeautifulSoup
from contextlib import asynccontextmanager
from fastapi import FastAPI, Request
from fastapi.responses import JSONResponse
from fastapi.middleware.cors import CORSMiddleware
from huggingface_hub import hf_hub_download
from llama_cpp import Llama
# ==========================================
# 1. CONFIGURATION & CORE SETUP
# ==========================================
# IMPORTANT: Set REAPERAI_SECRET as a Secret in your Space Settings
SECRET_KEY = os.environ.get("REAPERAI_SECRET", "jan30")
# Optimized for 16GB RAM: Smaller, heavily quantized model
MODEL_REPO = "bartowski/Qwen2.5-1.5B-Instruct-GGUF"
MODEL_FILE = "Qwen2.5-1.5B-Instruct-Q4_K_M.gguf" # Exact filename
chat_memory = {}
MAX_GLOBAL_USERS = 50
memory_lock = threading.Lock()
llm = None # Initialize as None
model_semaphore = threading.Semaphore(1) # Only 1 inference at a time
# ==========================================
# 2. MODEL LOADING (OPTIMIZED FOR SPACES)
# ==========================================
@asynccontextmanager
async def lifespan(app: FastAPI):
"""
Lifespan handler for FastAPI startup/shutdown.
Downloads and loads the model.
"""
print(f"--- [SYSTEM] Initializing ReaperAI on Hugging Face Space ---")
global llm
try:
# Step 1: Download model to cache (will use /tmp from Dockerfile)
print(f"--- [SYSTEM] Downloading model: {MODEL_REPO}/{MODEL_FILE} ---")
model_path = hf_hub_download(
repo_id=MODEL_REPO,
filename=MODEL_FILE,
cache_dir=os.getenv("HF_HOME", "/tmp")
)
# Step 2: Load with optimized settings for 2 vCPU / 16GB RAM
print(f"--- [SYSTEM] Loading model into RAM (this may take a moment) ---")
llm = Llama(
model_path=model_path,
n_ctx=1024, # Reduced for memory efficiency
n_threads=2, # Matches your 2 vCPUs
n_gpu_layers=0, # CPU only
verbose=False
)
print(f"--- [SYSTEM] Model loaded successfully. ReaperAI is ready. ---")
except Exception as e:
print(f"--- [CRITICAL ERROR] Model loading failed: {str(e)} ---")
llm = None # Ensure it's None if loading fails
yield # App runs here
# Cleanup on shutdown (optional)
if llm is not None:
del llm
gc.collect()
# ==========================================
# 3. FASTAPI APP INITIALIZATION
# ==========================================
app = FastAPI(
title="ReaperAI Secure Core",
description="AI Assistant with Web Search Capabilities",
version="2.0",
lifespan=lifespan
)
# CORS configuration
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_methods=["POST", "GET"],
allow_headers=["*"],
)
# ==========================================
# 4. AUTONOMOUS TOOLS (REFINED)
# ==========================================
def ddg_search(query):
"""Perform a DuckDuckGo search and return top 3 results."""
print(f"--- [TOOL] Searching Web: {query} ---")
try:
headers = {"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36"}
res = requests.get(
"https://html.duckduckgo.com/html/",
params={"q": query, "kl": "us-en"},
headers=headers,
timeout=8
)
res.raise_for_status()
soup = BeautifulSoup(res.text, "html.parser")
results = []
for r in soup.select(".result")[:3]: # Limit to 3 results
title_elem = r.select_one('.result__a')
snippet_elem = r.select_one('.result__snippet')
if title_elem and snippet_elem:
title = title_elem.get_text(strip=True)
snippet = snippet_elem.get_text(strip=True)[:200] # Truncate
results.append(f"• {title}: {snippet}")
if results:
return f"\n[REAL-TIME SEARCH RESULTS]:\n" + "\n".join(results) + "\n"
return ""
except Exception as e:
print(f"--- [TOOL ERROR] Search failed: {e} ---")
return ""
def jina_read(url):
"""Fetch and parse content from a URL using Jina Reader."""
print(f"--- [TOOL] Reading Source: {url} ---")
try:
# Clean the URL
url = url.strip()
if not url.startswith(('http://', 'https://')):
url = 'https://' + url
res = requests.get(
f"https://r.jina.ai/{url}",
headers={"x-respond-with": "text", "User-Agent": "ReaperAI/2.0"},
timeout=10
)
res.raise_for_status()
# Extract first 1200 chars for context
content = res.text[:1200].strip()
if content:
return f"\n[SOURCE CONTENT]:\n{content}\n"
return ""
except Exception as e:
print(f"--- [TOOL ERROR] URL read failed: {e} ---")
return ""
def fast_intent_detection(message):
"""Detect user intent from message."""
message_lower = message.lower()
# Check for URL
url_match = re.search(r"(https?://\S+)", message)
if url_match:
return "URL", url_match.group(1)
# Check for search keywords
search_keywords = [
"who is", "what is", "how to", "price of", "latest", "current",
"news", "today", "weather", "score", "stock", "update", "2024",
"define", "explain"
]
if any(keyword in message_lower for keyword in search_keywords):
return "SEARCH", message
return "CHAT", None
# ==========================================
# 5. SECURITY MIDDLEWARE & ENDPOINTS
# ==========================================
@app.middleware("http")
async def security_guard(request: Request, call_next):
"""Security middleware for API key validation."""
# Allow root endpoint without auth
if request.url.path == "/":
return await call_next(request)
# Check for API key in headers
if request.headers.get("x-reaperai-key") != SECRET_KEY:
return JSONResponse(
status_code=403,
content={"error": "ACCESS_DENIED", "message": "Invalid or missing API key"}
)
return await call_next(request)
@app.get("/")
async def root():
"""Root endpoint for health checks."""
status = "ready" if llm is not None else "loading"
return {
"status": status,
"service": "ReaperAI Secure Core",
"version": "2.0",
"model_loaded": llm is not None,
"endpoints": {"/chat": "POST", "/health": "GET"}
}
@app.get("/health")
async def health_check():
"""Health check endpoint for monitoring."""
return {
"status": "healthy" if llm is not None else "unhealthy",
"model": MODEL_REPO if llm is not None else None,
"memory_users": len(chat_memory),
"timestamp": datetime.now().isoformat()
}
@app.post("/chat")
async def chat_endpoint(request: Request):
"""Main chat endpoint."""
try:
payload = await request.json()
except:
return JSONResponse(
status_code=400,
content={"error": "INVALID_JSON", "response": "Request must be valid JSON"}
)
user_id = payload.get("userId", "default")
message = payload.get("message", "").strip()
current_date = datetime.now().strftime("%A, %B %d, %Y")
# Validate input
if not message:
return JSONResponse(
status_code=400,
content={"error": "EMPTY_MESSAGE", "response": "Message cannot be empty"}
)
# Check if model is loaded
if llm is None:
return JSONResponse(
status_code=503,
content={
"error": "MODEL_NOT_LOADED",
"response": "AI model is still initializing. Please try again in 30 seconds."
}
)
# Manage conversation history (thread-safe)
with memory_lock:
# Clean up old users if needed
if len(chat_memory) > MAX_GLOBAL_USERS:
oldest_user = next(iter(chat_memory))
del chat_memory[oldest_user]
# Get user's history (last 5 exchanges)
if user_id not in chat_memory:
chat_memory[user_id] = []
history = chat_memory[user_id][-5:]
# Determine intent and gather context
intent, data = fast_intent_detection(message)
context = ""
if intent == "URL":
context = jina_read(data)
elif intent == "SEARCH":
context = ddg_search(message)
# Build conversation messages
messages = [
{
"role": "system",
"content": f"""You are ReaperAI, a helpful and concise AI assistant.
Current Date: {current_date}
Instructions:
1. Be direct and informative
2. Use provided context when available
3. Keep responses under 300 words
4. If you don't know, say so
"""
}
]
# Add conversation history
for h in history:
messages.append({"role": "user", "content": h['u']})
messages.append({"role": "assistant", "content": h['a']})
# Add current query with context
if context:
final_query = f"Context:\n{context}\n\nUser Query: {message}"
else:
final_query = message
messages.append({"role": "user", "content": final_query})
# Generate response (thread-safe)
with model_semaphore:
try:
response = llm.create_chat_completion(
messages=messages,
max_tokens=400, # Limit response length
temperature=0.7,
stop=["###", "User:", "Assistant:"]
)
ai_response = response["choices"][0]["message"]["content"].strip()
except Exception as e:
print(f"--- [INFERENCE ERROR] {str(e)} ---")
ai_response = f"I encountered an error processing your request. Please try again."
# Update conversation history (thread-safe)
with memory_lock:
chat_memory[user_id].append({"u": message, "a": ai_response})
# Keep last 8 exchanges per user
chat_memory[user_id] = chat_memory[user_id][-8:]
# Clean up
gc.collect()
return {
"intent": intent,
"response": ai_response,
"context_used": bool(context),
"user_id": user_id,
"timestamp": datetime.now().isoformat()
}
# ==========================================
# 6. MAIN EXECUTION
# ==========================================
if __name__ == "__main__":
import uvicorn
print("--- [SYSTEM] Starting ReaperAI Server ---")
uvicorn.run(
"main:app",
host="0.0.0.0",
port=7860,
reload=False, # Disable reload in production
timeout_keep_alive=60,
access_log=True
) |