api / main.py
infinityonline's picture
Update main.py
40bf578 verified
import os
import uuid
import time
import asyncio
import threading
import json
import re
from typing import Optional
from fastapi import FastAPI, Header, HTTPException, Request
from fastapi.responses import JSONResponse
API_SECRET_KEY = os.getenv("API_SECRET_KEY", "2026-2026")
class AsyncBrowserThread(threading.Thread):
def __init__(self):
super().__init__(daemon=True)
self.loop = asyncio.new_event_loop()
self.ready_event = threading.Event()
self.browser = None
self.playwright = None
def run(self):
asyncio.set_event_loop(self.loop)
self.loop.run_until_complete(self._start_browser())
self.ready_event.set()
print("[LITE-SERVER] Browser is ready!")
self.loop.run_forever()
async def _start_browser(self):
from playwright.async_api import async_playwright
print("[LITE-SERVER] Starting Chrome...")
self.playwright = await async_playwright().start()
self.browser = await self.playwright.chromium.launch(
headless=True,
channel="chrome",
args=[
'--disable-blink-features=AutomationControlled',
'--no-sandbox',
'--disable-gpu',
'--disable-dev-shm-usage',
'--disable-setuid-sandbox',
'--single-process',
'--no-zygote',
]
)
print("[LITE-SERVER] Chrome launched!")
async def _talk_to_chatgpt(self, prompt: str):
context = await self.browser.new_context(
user_agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0.0.0 Safari/537.36",
viewport={'width': 1920, 'height': 1080}
)
await context.add_init_script("Object.defineProperty(navigator, 'webdriver', {get: () => undefined})")
page = await context.new_page()
try:
page.set_default_timeout(120000)
await page.goto("https://chatgpt.com/", wait_until="domcontentloaded")
await page.wait_for_selector('#prompt-textarea', timeout=60000)
await page.fill('#prompt-textarea', prompt)
await asyncio.sleep(0.5)
await page.press('#prompt-textarea', 'Enter')
await page.wait_for_selector('[data-message-author-role="assistant"]', timeout=120000)
last_text = ""
unchanged_count = 0
while unchanged_count < 4:
messages = await page.query_selector_all('[data-message-author-role="assistant"]')
if messages:
current_text = await messages[-1].inner_text()
if current_text == last_text and current_text.strip() != "":
unchanged_count += 1
else:
last_text = current_text
unchanged_count = 0
await asyncio.sleep(0.5)
return last_text.strip()
except Exception as e:
print(f"[LITE-SERVER] Error: {e}")
raise e
finally:
await page.close()
await context.close()
def process_request(self, prompt: str):
if not self.ready_event.wait(timeout=60):
raise Exception("Error From Browser")
future = asyncio.run_coroutine_threadsafe(self._talk_to_chatgpt(prompt), self.loop)
return future.result(timeout=120)
browser_engine = AsyncBrowserThread()
browser_engine.start()
# ====================================================================
# Smart Prompt Builder
# ====================================================================
def format_prompt(messages, tools=None):
parts = []
system_parts = []
has_tool_results = False
user_question = ""
for msg in messages:
role = msg.get("role", "")
msg_type = msg.get("type", "")
content = msg.get("content", "")
if isinstance(content, list):
text_parts = []
for item in content:
if isinstance(item, dict):
text_parts.append(item.get("text", item.get("content", str(item))))
else:
text_parts.append(str(item))
content = "\n".join(text_parts)
if role == "system":
system_parts.append(content)
elif role == "tool":
has_tool_results = True
tool_name = msg.get("name", "tool")
parts.append(f"[TOOL RESULT from '{tool_name}']:\n{content}")
elif msg_type == "function_call_output":
has_tool_results = True
call_id = msg.get("call_id", "")
output_content = msg.get("output", content)
parts.append(f"[TOOL RESULT (call_id: {call_id})]:\n{output_content}")
elif msg_type == "function_call":
func_name = msg.get("name", "?")
func_args = msg.get("arguments", "{}")
parts.append(f"[PREVIOUS TOOL CALL: Called '{func_name}' with arguments: {func_args}]")
elif role == "assistant":
assistant_content = content if content else ""
tool_calls_in_msg = msg.get("tool_calls", [])
if tool_calls_in_msg:
tc_descriptions = []
for tc in tool_calls_in_msg:
func = tc.get("function", {})
tc_descriptions.append(f"Called '{func.get('name', '?')}' with: {func.get('arguments', '{}')}")
assistant_content += "\n[Previous tool calls: " + "; ".join(tc_descriptions) + "]"
if assistant_content.strip():
parts.append(f"[Assistant]: {assistant_content}")
elif role == "user" or (msg_type == "message" and role != "system"):
user_question = content
parts.append(content)
has_tool_results = False
elif content:
parts.append(content)
final = ""
if system_parts:
if tools and not has_tool_results:
final += "=== YOUR ROLE ===\n"
final += "\n\n".join(system_parts)
final += "\n=== END OF ROLE ===\n\n"
else:
final += "=== SYSTEM INSTRUCTIONS (FOLLOW STRICTLY) ===\n"
final += "\n\n".join(system_parts)
final += "\n=== END OF INSTRUCTIONS ===\n\n"
if tools and not has_tool_results:
final += format_tools_instruction(tools, user_question)
if has_tool_results:
final += "=== CONTEXT FROM TOOLS ===\n"
final += "The following information was retrieved by the tools you requested.\n"
final += "Use ONLY this information to answer the user's question.\n\n"
if parts:
final += "\n".join(parts)
if has_tool_results:
final += "\n\n=== INSTRUCTION ===\n"
final += "Now answer the user's question based ONLY on the tool results above.\n"
return final
def format_tools_instruction(tools, user_question=""):
instruction = "\n=== MANDATORY TOOL USAGE ===\n"
instruction += "You MUST use one of the tools below to answer this question.\n"
instruction += "Do NOT answer directly. Do NOT say you don't have information.\n"
instruction += "You MUST respond with ONLY a JSON object to call the tool.\n\n"
instruction += "RESPONSE FORMAT - respond with ONLY this JSON, nothing else:\n"
instruction += '{"tool_calls": [{"name": "TOOL_NAME", "arguments": {"param": "value"}}]}\n\n'
instruction += "RULES:\n"
instruction += "- Your ENTIRE response must be valid JSON only\n"
instruction += "- No markdown, no code blocks, no explanation\n"
instruction += "- No text before or after the JSON\n\n"
instruction += "Available tools:\n\n"
for tool in tools:
func = tool.get("function", tool)
name = func.get("name", "unknown")
desc = func.get("description", "No description")
params = func.get("parameters", {})
instruction += f"Tool: {name}\n"
instruction += f"Description: {desc}\n"
if params.get("properties"):
instruction += "Parameters:\n"
required_params = params.get("required", [])
for param_name, param_info in params["properties"].items():
param_type = param_info.get("type", "string")
param_desc = param_info.get("description", "")
is_required = "required" if param_name in required_params else "optional"
instruction += f" - {param_name} ({param_type}, {is_required}): {param_desc}\n"
instruction += "\n"
instruction += "=== END OF TOOLS ===\n\n"
first_tool = tools[0] if tools else {}
first_func = first_tool.get("function", first_tool)
first_name = first_func.get("name", "tool")
instruction += f'EXAMPLE: If the user asks a question, respond with:\n'
instruction += '{"tool_calls": [{"name": "' + first_name + '", "arguments": {"input": "the user question here"}}]}\n\n'
instruction += "Now respond with the JSON to call the appropriate tool:\n\n"
return instruction
def parse_tool_calls(response_text):
cleaned = response_text.strip()
if "```" in cleaned:
code_block_match = re.search(r'```(?:json)?\s*\n?(.*?)\n?\s*```', cleaned, re.DOTALL)
if code_block_match:
cleaned = code_block_match.group(1).strip()
json_candidates = [cleaned]
json_match = re.search(r'\{[\s\S]*"tool_calls"[\s\S]*\}', cleaned)
if json_match:
json_candidates.append(json_match.group(0))
for candidate in json_candidates:
try:
parsed = json.loads(candidate)
if isinstance(parsed, dict) and "tool_calls" in parsed:
raw_calls = parsed["tool_calls"]
if isinstance(raw_calls, list) and len(raw_calls) > 0:
formatted_calls = []
for call in raw_calls:
tool_name = call.get("name", "")
arguments = call.get("arguments", {})
if isinstance(arguments, dict):
arguments_str = json.dumps(arguments, ensure_ascii=False)
else:
arguments_str = str(arguments)
formatted_calls.append({
"id": f"call_{uuid.uuid4().hex[:24]}",
"type": "function",
"function": {
"name": tool_name,
"arguments": arguments_str
}
})
return formatted_calls
except (json.JSONDecodeError, TypeError, KeyError):
continue
return None
# ====================================================================
# FastAPI App
# ====================================================================
app = FastAPI(title="mse_ai_api for n8n")
@app.post("/v1/chat/completions")
async def chat_completions(request: Request):
try:
data = await request.json()
except Exception:
return JSONResponse(status_code=400, content={"error": {"message": "Invalid JSON payload"}})
authorization = request.headers.get("authorization", "")
if not authorization or authorization.replace("Bearer ", "").strip() != API_SECRET_KEY:
return JSONResponse(status_code=401, content={"error": {"message": "Invalid API Key"}})
messages = data.get("messages", [])
if not messages:
return JSONResponse(status_code=400, content={"error": {"message": "messages field is required"}})
try:
tools = data.get("tools", None)
prompt = format_prompt(messages, tools=tools)
start_time = time.time()
print(f"[LITE-SERVER] Processing request ({len(prompt)} chars)")
response_text = browser_engine.process_request(prompt)
p_tokens = len(prompt.split())
c_tokens = len(response_text.split())
tool_calls = None
if tools:
tool_calls = parse_tool_calls(response_text)
if tool_calls:
return {
"id": f"chatcmpl-{uuid.uuid4().hex[:29]}",
"object": "chat.completion",
"created": int(start_time),
"model": data.get("model", "gpt-4o-mini"),
"choices": [{"index": 0, "message": {"role": "assistant", "content": None,
"tool_calls": tool_calls}, "finish_reason": "tool_calls"}],
"usage": {"prompt_tokens": p_tokens, "completion_tokens": c_tokens,
"total_tokens": p_tokens + c_tokens}
}
else:
return {
"id": f"chatcmpl-{uuid.uuid4().hex[:29]}",
"object": "chat.completion",
"created": int(start_time),
"model": data.get("model", "gpt-4o-mini"),
"choices": [{"index": 0, "message": {"role": "assistant", "content": response_text},
"finish_reason": "stop"}],
"usage": {"prompt_tokens": p_tokens, "completion_tokens": c_tokens,
"total_tokens": p_tokens + c_tokens}
}
except Exception as e:
return JSONResponse(status_code=500, content={"error": str(e)})
@app.post("/v1/responses")
async def responses(request: Request):
try:
data = await request.json()
except Exception:
return JSONResponse(status_code=400, content={"error": {"message": "Invalid JSON payload"}})
authorization = request.headers.get("authorization", "")
if not authorization or authorization.replace("Bearer ", "").strip() != API_SECRET_KEY:
return JSONResponse(status_code=401, content={"error": {"message": "Invalid API Key"}})
input_data = data.get("input", "")
if isinstance(input_data, str):
messages = [{"role": "user", "content": input_data}]
elif isinstance(input_data, list):
messages = input_data
else:
messages = data.get("messages", [])
if not messages:
return JSONResponse(status_code=400, content={"error": {"message": "input field is required"}})
try:
tools = data.get("tools", None)
instructions = data.get("instructions", "")
if instructions:
messages.insert(0, {"role": "system", "content": instructions})
prompt = format_prompt(messages, tools=tools)
start_time = time.time()
response_text = browser_engine.process_request(prompt)
p_tokens = len(prompt.split())
c_tokens = len(response_text.split())
tool_calls = None
if tools:
tool_calls = parse_tool_calls(response_text)
if tool_calls:
output_items = []
for tc in tool_calls:
output_items.append({
"type": "function_call",
"id": tc["id"],
"call_id": tc["id"],
"name": tc["function"]["name"],
"arguments": tc["function"]["arguments"],
"status": "completed"
})
return {
"id": f"resp-{uuid.uuid4().hex[:29]}",
"object": "response",
"created_at": int(start_time),
"model": data.get("model", "gpt-4o-mini"),
"status": "completed",
"output": output_items,
"usage": {"input_tokens": p_tokens, "output_tokens": c_tokens,
"total_tokens": p_tokens + c_tokens}
}
else:
return {
"id": f"resp-{uuid.uuid4().hex[:29]}",
"object": "response",
"created_at": int(start_time),
"model": data.get("model", "gpt-4o-mini"),
"status": "completed",
"output": [{"type": "message", "role": "assistant",
"content": [{"type": "output_text", "text": response_text}]}],
"usage": {"input_tokens": p_tokens, "output_tokens": c_tokens,
"total_tokens": p_tokens + c_tokens}
}
except Exception as e:
return JSONResponse(status_code=500, content={"error": str(e)})
@app.get("/v1/models")
async def list_models():
return {"object": "list",
"data": [{"id": "gpt-4o-mini", "object": "model", "owned_by": "mse_ai_api"}]}
# ── تغيير 1: إضافة /health endpoint (مطلوب لـ HuggingFace healthcheck)
@app.get("/health")
@app.get("/")
async def health_check():
return {"status": "running", "message": "mse_ai_api Server is active!"}
if __name__ == "__main__":
import uvicorn
# ── تغيير 2: port 7777 → 7860 (مطلوب لـ HuggingFace)
uvicorn.run(app, host="0.0.0.0", port=7860)