import base64 from contextlib import asynccontextmanager import logging import time import uuid import os import re import asyncio import multiprocessing import hashlib import json import psutil from typing import List, Optional, Dict, Any, TypeVar, Generic from urllib.parse import urlparse, parse_qs, urlencode, urlunparse from threading import Lock # --- FastAPI & Core --- from fastapi import FastAPI, HTTPException, Depends from fastapi.encoders import jsonable_encoder from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials from fastapi.middleware.cors import CORSMiddleware from starlette.concurrency import run_in_threadpool from anyio import to_thread from dotenv import load_dotenv from pydantic import BaseModel # --- Async Database Drivers --- import asyncpg import asyncmy from asyncmy.cursors import DictCursor # --- Database Drivers (Kept for compatibility/types if needed, but unused in SQL routes) --- from bson import ObjectId import mysql.connector import psycopg2 from psycopg2.extras import RealDictCursor import uvicorn # --- Existing Services --- from csv_analysis_service import execute_analysis_logic from csv_chart_service import execute_python_code from csv_metadata_service import CsvDataRequest, CsvInfoRequest, CsvInfoResponse, PythonExecutionRequest, PythonExecutionResponse, execute_python_logic, get_csv_basic_info, get_robust_csv_rows from mongo_service import execute_mongo_operation, parse_query_input from pydantic_csv_analysis_model import AnalysisRequest, AnalysisResponse from pydantic_csv_charts_model import ChartExecutionPayload, ChartExecutionResponse from pydantic_mongo_executor_model import ExecutorPayload, ExecutorResponse from report_service import FileBoxProps, ReportRequest, execute_report_generation from supabase_service import upload_bytes_to_supabase # --- Configuration & Setup --- load_dotenv() logging.basicConfig( format="%(asctime)s - %(name)s - %(levelname)s - %(message)s", level=logging.INFO ) logger = logging.getLogger("API_Controller") # ============================================================================== # ASYNC CONNECTION POOL MANAGER # ============================================================================== class AsyncPoolManager: def __init__(self): self._pg_pools: Dict[str, asyncpg.Pool] = {} self._mysql_pools: Dict[str, asyncmy.Pool] = {} self._lock = asyncio.Lock() async def get_pg_pool(self, db_url: str) -> asyncpg.Pool: async with self._lock: if db_url not in self._pg_pools: logger.info(f"Creating new AsyncPG pool for: {db_url[:20]}...") try: # asyncpg handles Keepalives and SSL automatically better than psycopg2 pool = await asyncpg.create_pool( dsn=db_url, min_size=1, max_size=20, max_inactive_connection_lifetime=300, command_timeout=60 ) self._pg_pools[db_url] = pool except Exception as e: logger.error(f"Failed to create PG pool: {e}") raise e return self._pg_pools[db_url] async def get_mysql_pool(self, db_url: str) -> asyncmy.Pool: async with self._lock: if db_url not in self._mysql_pools: logger.info(f"Creating new AsyncMy pool for: {db_url[:20]}...") parsed = urlparse(db_url) # Check for SSL requirement in query params qs = parse_qs(parsed.query) # AsyncMy often auto-negotiates SSL, but we can enforce it if needed based on qs try: pool = await asyncmy.create_pool( user=parsed.username, password=parsed.password, host=parsed.hostname, port=parsed.port or 3306, db=parsed.path.lstrip("/"), minsize=1, maxsize=20, autocommit=True, pool_recycle=280, # Recycle before default 8hr timeout to prevent "Gone Away" ) self._mysql_pools[db_url] = pool except Exception as e: logger.error(f"Failed to create MySQL pool: {e}") raise e return self._mysql_pools[db_url] async def close_all(self): logger.info("Closing all async database pools...") for url, pool in self._pg_pools.items(): await pool.close() for url, pool in self._mysql_pools.items(): pool.close() await pool.wait_closed() # Initialize Global Async Manager async_pool_manager = AsyncPoolManager() def get_dynamic_thread_limit(): """Calculates a safe thread limit based on available RAM.""" try: total_ram_bytes = psutil.virtual_memory().total total_cores = multiprocessing.cpu_count() num_workers = max(1, total_cores - 2) ram_per_worker = total_ram_bytes / num_workers safe_ram_pool = ram_per_worker * 0.70 BYTES_PER_THREAD = 8 * 1024 * 1024 calculated_limit = int(safe_ram_pool / BYTES_PER_THREAD) final_limit = max(100, min(calculated_limit, 3000)) logger.info(f"Dynamic Limit Config: {total_ram_bytes/(1024**3):.2f}GB RAM / {num_workers} Workers. Limit: {final_limit}") return final_limit except Exception as e: logger.warning(f"Failed to calculate dynamic threads ({e}). Fallback to 1000.") return 1000 @asynccontextmanager async def lifespan(app: FastAPI): # Startup safe_limit = get_dynamic_thread_limit() to_thread.current_default_thread_limiter().total_tokens = safe_limit logger.info(f"Worker Process Started: Thread pool capacity set to {safe_limit}. Async Drivers Ready.") yield # Shutdown await async_pool_manager.close_all() app = FastAPI(title="Unified Data Executor API", lifespan=lifespan) # ============================================================================== # MODELS & MIDDLEWARE # ============================================================================== T = TypeVar("T") class BatchRequest(BaseModel, Generic[T]): requests: List[T] class BatchResponse(BaseModel, Generic[T]): responses: List[T] CHART_DIR = "generated_charts" os.makedirs(CHART_DIR, exist_ok=True) origins_env = os.getenv("ALLOWED_ORIGINS", "*") ORIGINS = [origin.strip() for origin in origins_env.split(",")] app.add_middleware( CORSMiddleware, allow_origins=ORIGINS, allow_credentials=True, allow_methods=["*"], allow_headers=["*"], ) security = HTTPBearer() API_SECRET_TOKEN = os.getenv("API_BEARER_TOKEN") async def validate_token(credentials: HTTPAuthorizationCredentials = Depends(security)): if credentials.credentials != API_SECRET_TOKEN: raise HTTPException(status_code=403, detail="Invalid Authentication Token") return credentials.credentials # ============================================================================== # REQUEST COALESCER # ============================================================================== class RequestCoalescer: def __init__(self): self._active_requests: Dict[str, asyncio.Future] = {} self._lock = asyncio.Lock() def _generate_key(self, prefix: str, data: dict) -> str: json_str = json.dumps(data, sort_keys=True, default=str) raw_str = f"{prefix}:{json_str}" return hashlib.md5(raw_str.encode()).hexdigest() async def execute(self, prefix: str, unique_params: dict, func, *args, **kwargs): key = self._generate_key(prefix, unique_params) async with self._lock: if key in self._active_requests: return await self._active_requests[key] future = asyncio.get_running_loop().create_future() self._active_requests[key] = future try: # Works for both async functions and run_in_threadpool result = await func(*args, **kwargs) if not future.done(): future.set_result(result) return result except Exception as e: if not future.done(): future.set_exception(e) raise e finally: async with self._lock: if key in self._active_requests: del self._active_requests[key] coalescer = RequestCoalescer() # ============================================================================== # ASYNC DB LOGIC # ============================================================================== def is_aggregate_query(query: str) -> bool: query_lower = query.lower() aggregate_patterns = [r'\bcount\s*\(', r'\bsum\s*\(', r'\bavg\s*\(', r'\bmin\s*\(', r'\bmax\s*\(', r'\bgroup\s+by\b', r'\bdistinct\b', r'\bhaving\b'] for pattern in aggregate_patterns: if re.search(pattern, query_lower): return True return False def normalize_mysql_uri(uri: str) -> str: try: parsed_uri = urlparse(uri) query_params = parse_qs(parsed_uri.query) query_params.pop('ssl-mode', None) new_query = urlencode(query_params, doseq=True) parsed_uri = parsed_uri._replace(query=new_query) return urlunparse(parsed_uri) except Exception: return uri def normalize_postgres_uri(uri: str) -> str: try: parsed_uri = urlparse(uri) if parsed_uri.scheme == 'postgres': parsed_uri = parsed_uri._replace(scheme='postgresql') return urlunparse(parsed_uri) except Exception: return uri # --- ASYNC MYSQL EXECUTOR --- async def _execute_async_mysql(db_url: str, sql_query: str, max_rows: int = 20, limited: bool = False) -> dict: start_time = time.time() try: # Get pool pool = await async_pool_manager.get_mysql_pool(db_url) async with pool.acquire() as conn: async with conn.cursor(cursor=DictCursor) as cursor: clean_query = sql_query.strip() query_lower = clean_query.lower() if not query_lower.startswith("select"): await cursor.execute(clean_query) return {"success": True, "message": "Query executed successfully (Non-SELECT).", "executionTime": time.time() - start_time} final_query = clean_query is_aggregate = is_aggregate_query(clean_query) is_limited_result = False message = "" if not limited: message = f"Raw query executed." else: if is_aggregate: message = f"Aggregate query completed." else: final_query = clean_query.rstrip(';').strip() if not re.search(r'\blimit\s+\d+', query_lower): final_query = f"{final_query} LIMIT {max_rows}" is_limited_result = True message = f"Showing first {max_rows} rows." await cursor.execute(final_query) results = await cursor.fetchall() # Check actual length if is_limited_result and len(results) < max_rows: is_limited_result = False message = f"Returned {len(results)} rows." elif is_limited_result: message = f"Showing first {len(results)} rows." columns = [col[0] for col in cursor.description] if cursor.description else [] return { "success": True, "results": jsonable_encoder(results), "columns": columns, "rowCount": len(results), "executionTime": time.time() - start_time, "is_aggregate": is_aggregate, "limited": is_limited_result, "message": message, "error": None } except Exception as e: return {"success": False, "error": str(e), "executionTime": 0.0} # --- ASYNC POSTGRES EXECUTOR --- async def _execute_async_postgres(db_url: str, sql_query: str, max_rows: int = 20, limited: bool = False) -> dict: start_time = time.time() try: pool = await async_pool_manager.get_pg_pool(db_url) async with pool.acquire() as conn: clean_query = sql_query.strip() query_lower = clean_query.lower() if not query_lower.startswith(("select", "show", "explain", "with")): await conn.execute(clean_query) return {"success": True, "message": "Query executed successfully (Non-SELECT).", "executionTime": time.time() - start_time} final_query = clean_query is_aggregate = is_aggregate_query(clean_query) is_limited_result = False message = "" if not limited: message = f"Raw query executed." else: if is_aggregate: message = f"Aggregate query completed." else: final_query = clean_query.rstrip(';').strip() if not re.search(r'\blimit\s+\d+', query_lower): final_query = f"{final_query} LIMIT {max_rows}" is_limited_result = True message = f"Showing first {max_rows} rows." # Fetch results records = await conn.fetch(final_query) results = [dict(r) for r in records] if is_limited_result and len(results) < max_rows: is_limited_result = False message = f"Returned {len(results)} rows." elif is_limited_result: message = f"Showing first {len(results)} rows." columns = list(results[0].keys()) if results else [] return { "success": True, "results": jsonable_encoder(results, custom_encoder={uuid.UUID: str, ObjectId: str}), "columns": columns, "rowCount": len(results), "executionTime": time.time() - start_time, "is_aggregate": is_aggregate, "limited": is_limited_result, "message": message, "error": None } except Exception as e: return {"success": False, "error": str(e), "executionTime": 0.0} # ============================================================================== # ROUTES # ============================================================================== class SqlQueryRequest(BaseModel): database_url: str sql_query: str limit_rows: Optional[int] = 20 limited: bool = False class SqlQueryResponse(BaseModel): success: bool results: Optional[List[Dict[str, Any]]] = None columns: Optional[List[str]] = None rowCount: Optional[int] = 0 executionTime: Optional[float] = 0.0 error: Optional[str] = None request_id: str is_aggregate: bool = False limited: bool = False message: Optional[str] = None class PgQueryRequest(BaseModel): database_url: str sql_query: str limit_rows: Optional[int] = 20 limited: bool = False class PgQueryResponse(BaseModel): success: bool results: Optional[List[Dict[str, Any]]] = None columns: Optional[List[str]] = None rowCount: Optional[int] = 0 executionTime: Optional[float] = 0.0 error: Optional[str] = None request_id: str is_aggregate: bool = False limited: bool = False message: Optional[str] = None @app.post("/api/execute_sql_query", response_model=SqlQueryResponse) async def execute_mysql_endpoint(query: SqlQueryRequest, token: str = Depends(validate_token)): request_id = str(uuid.uuid4())[:8] try: normalized_url = normalize_mysql_uri(query.database_url) limit_val = query.limit_rows if query.limit_rows is not None else 20 unique_params = { "db": normalized_url, "q": query.sql_query, "l": limit_val, "lim": query.limited } # Direct Async Call (No Threadpool) result_dict = await coalescer.execute( "mysql", unique_params, _execute_async_mysql, db_url=normalized_url, sql_query=query.sql_query, max_rows=limit_val, limited=query.limited ) final_result = result_dict.copy() final_result["request_id"] = request_id return SqlQueryResponse(**final_result) except Exception as e: raise HTTPException(status_code=500, detail={"success": False, "error": str(e), "request_id": request_id}) @app.post("/api/execute_postgres_query", response_model=PgQueryResponse) async def execute_postgres_endpoint(query: PgQueryRequest, token: str = Depends(validate_token)): request_id = str(uuid.uuid4())[:8] try: clean_url = normalize_postgres_uri(query.database_url) limit_val = query.limit_rows if query.limit_rows is not None else 20 unique_params = { "db": clean_url, "q": query.sql_query, "l": limit_val, "lim": query.limited } # Direct Async Call (No Threadpool) result_dict = await coalescer.execute( "postgres", unique_params, _execute_async_postgres, db_url=clean_url, sql_query=query.sql_query, max_rows=limit_val, limited=query.limited ) final_result = result_dict.copy() final_result["request_id"] = request_id return PgQueryResponse(**final_result) except Exception as e: raise HTTPException(status_code=500, detail={"success": False, "error": str(e), "request_id": request_id}) @app.post("/api/execute_mongo", response_model=ExecutorResponse) async def execute_mongo_endpoint(payload: ExecutorPayload, token: str = Depends(validate_token)): request_id = str(uuid.uuid4())[:8] start_time = time.time() try: parsed_query = parse_query_input(payload.generated_query) unique_params = { "uri": payload.mongo_uri, "db": payload.db_name, "col": payload.collection_name, "q": parsed_query, "lim": payload.limited, "lrows": payload.limit_rows } # Mongo is Sync via PyMongo -> Uses run_in_threadpool result_data = await coalescer.execute( "mongo", unique_params, run_in_threadpool, execute_mongo_operation, mongo_uri=payload.mongo_uri, db_name=payload.db_name, collection_name=payload.collection_name, query=parsed_query, limited=payload.limited, limit_rows=payload.limit_rows ) return ExecutorResponse(status="success", count=len(result_data), data=jsonable_encoder(result_data, custom_encoder={ObjectId: str}), duration_seconds=round(time.time() - start_time, 4), request_id=request_id) except Exception as e: raise HTTPException(status_code=500, detail=str(e)) @app.post("/api/execute_chart", response_model=ChartExecutionResponse) async def execute_chart_endpoint(payload: ChartExecutionPayload, token: str = Depends(validate_token)): request_id = str(uuid.uuid4())[:8] try: image_bytes, error_msg, logs = await run_in_threadpool(execute_python_code, code=payload.code, csv_url=payload.csv_url) if error_msg: return ChartExecutionResponse(status="error", error=error_msg, output_log=logs, request_id=request_id) if payload.return_base64: base64_str = base64.b64encode(image_bytes).decode('utf-8') return ChartExecutionResponse(status="success", base64_image=base64_str, output_log=logs, request_id=request_id) else: unique_name = f"{uuid.uuid4()}.png" public_url = await run_in_threadpool(upload_bytes_to_supabase, image_bytes=image_bytes, file_name=unique_name, chat_id=payload.chat_id) return ChartExecutionResponse(status="success", image_url=public_url, output_log=logs, request_id=request_id) except Exception as e: raise HTTPException(status_code=500, detail=str(e)) @app.post("/api/execute_csv_analysis", response_model=AnalysisResponse) async def execute_analysis_endpoint(payload: AnalysisRequest, token: str = Depends(validate_token)): request_id = str(uuid.uuid4())[:8] try: result = await run_in_threadpool(execute_analysis_logic, code=payload.code, csv_url=payload.csv_url) return AnalysisResponse(success=result["success"], output_log=result["output_log"], results=result["results"], error=result["error"], request_id=request_id) except Exception as e: raise HTTPException(status_code=500, detail=str(e)) @app.post("/api/generate_report", response_model=FileBoxProps) async def generate_report_endpoint(payload: ReportRequest, token: str = Depends(validate_token)): try: result = await execute_report_generation(code=payload.code, csv_url=payload.csv_url, chat_id=payload.chat_id) return result except Exception as e: raise HTTPException(status_code=500, detail=str(e)) @app.post("/api/get_csv_info", response_model=CsvInfoResponse) async def get_csv_info_endpoint(payload: CsvInfoRequest, token: str = Depends(validate_token)): request_id = str(uuid.uuid4())[:8] start_time = time.time() try: info_result = await run_in_threadpool(get_csv_basic_info, csv_path=payload.csv_url) if "error" in info_result: return CsvInfoResponse(success=False, error=info_result["error"], request_id=request_id, duration=time.time() - start_time) return CsvInfoResponse(success=True, data=info_result, request_id=request_id, duration=time.time() - start_time) except Exception as e: raise HTTPException(status_code=500, detail={"success": False, "error": str(e), "request_id": request_id}) @app.post("/api/csv_data") async def get_csv_data_endpoint(payload: CsvDataRequest, token: str = Depends(validate_token)): try: result = await run_in_threadpool(get_robust_csv_rows, csv_url=payload.csv_url) if isinstance(result, dict) and "error" in result: raise HTTPException(status_code=400, detail=result["error"]) return result except Exception as e: raise HTTPException(status_code=500, detail=f"Internal server error: {str(e)}") @app.post("/api/execute_python", response_model=PythonExecutionResponse) async def execute_python_endpoint(payload: PythonExecutionRequest, token: str = Depends(validate_token)): request_id = str(uuid.uuid4())[:8] try: execution_result = await run_in_threadpool(execute_python_logic, code=payload.code, custom_context=payload.context) return PythonExecutionResponse(success=execution_result['error'] is None, output=execution_result['output'], result=jsonable_encoder(execution_result['result']), isStructured=execution_result['isStructured'], error=execution_result['error'], request_id=request_id) except Exception as e: raise HTTPException(status_code=500, detail={"success": False, "error": str(e), "request_id": request_id}) # --- Batch Handlers --- async def batch_parallel_handler(func, requests: List[Any], token: str): tasks = [func(req, token) for req in requests] results = await asyncio.gather(*tasks, return_exceptions=True) return [res if not isinstance(res, Exception) else {"success": False, "error": str(res)} for res in results] @app.post("/api/batch/execute_sql_query", response_model=BatchResponse[SqlQueryResponse]) async def batch_execute_sql(payload: BatchRequest[SqlQueryRequest], token: str = Depends(validate_token)): responses = await batch_parallel_handler(execute_mysql_endpoint, payload.requests, token) return BatchResponse(responses=responses) @app.post("/api/batch/execute_postgres_query", response_model=BatchResponse[PgQueryResponse]) async def batch_execute_pg(payload: BatchRequest[PgQueryRequest], token: str = Depends(validate_token)): responses = await batch_parallel_handler(execute_postgres_endpoint, payload.requests, token) return BatchResponse(responses=responses) @app.post("/api/batch/execute_mongo", response_model=BatchResponse[ExecutorResponse]) async def batch_execute_mongo(payload: BatchRequest[ExecutorPayload], token: str = Depends(validate_token)): responses = await batch_parallel_handler(execute_mongo_endpoint, payload.requests, token) return BatchResponse(responses=responses) # --- Test Endpoint --- class TestCalcRequest(BaseModel): value: int def _heavy_calculation_task(value: int) -> int: time.sleep(0.1) return value * 2 @app.post("/api/test/parallel_calc_no_auth") async def test_parallel_calc_endpoint(payload: TestCalcRequest): try: result = await run_in_threadpool(_heavy_calculation_task, value=payload.value) return {"success": True, "result": result, "process_id": os.getpid()} except Exception as e: raise HTTPException(status_code=500, detail=str(e)) @app.get("/") async def root(): return {"message": "Python Code Execution Server is running"} @app.get("/ping") async def ping(): return {"message": "I am alive!"} if __name__ == "__main__": host = os.getenv("HOST", "0.0.0.0") port = int(os.getenv("PORT", 7860)) num_workers = max(1, multiprocessing.cpu_count() - 2) print(f"Starting production server on {host}:{port} with {num_workers} workers...") uvicorn.run("controller:app", host=host, port=port, workers=num_workers, loop="asyncio")