Files
gh-christianlouis-docuelevate/app/tasks/watch_folder_tasks.py
T

567 lines
21 KiB
Python

#!/usr/bin/env python3
"""
Watch Folder Ingestion Tasks
Periodically scans configured directories (local filesystem, FTP, SFTP) for new files
and enqueues them for document processing.
Supported sources:
- Local filesystem paths (works with any mounted path: SMB/CIFS, NFS, etc.)
- FTP server directories (uses existing FTP connection settings)
- SFTP server directories (uses existing SFTP connection settings)
"""
import ftplib # nosec B402 - FTP usage is intentional for legacy server support
import json
import logging
import os
import tempfile
from datetime import datetime, timedelta, timezone
import redis
from celery import shared_task
from app.config import settings
from app.tasks.convert_to_pdf import convert_to_pdf
from app.tasks.process_document import process_document
from app.utils.allowed_types import ALLOWED_EXTENSIONS, ALLOWED_MIME_TYPES
logger = logging.getLogger(__name__)
redis_client = redis.StrictRedis.from_url(settings.redis_url, decode_responses=True)
WATCH_FOLDER_LOCK_KEY = "watch_folder_lock"
WATCH_FOLDER_LOCK_EXPIRE = 300 # 5 minutes
# Cache file for tracking already-ingested files (local watch folders)
WATCH_FOLDER_CACHE_FILE = os.path.join(settings.workdir, "watch_folder_processed.json")
# Cache file for tracking already-ingested files (FTP watch folder)
FTP_INGEST_CACHE_FILE = os.path.join(settings.workdir, "ftp_ingest_processed.json")
# Cache file for tracking already-ingested files (SFTP watch folder)
SFTP_INGEST_CACHE_FILE = os.path.join(settings.workdir, "sftp_ingest_processed.json")
_CACHE_RETENTION_DAYS = 30
# ---------------------------------------------------------------------------
# Locking helpers
# ---------------------------------------------------------------------------
def _acquire_lock(lock_key: str, expire: int = WATCH_FOLDER_LOCK_EXPIRE) -> bool:
"""Acquire a Redis-based distributed lock. Returns True if acquired."""
acquired = redis_client.setnx(lock_key, "locked")
if acquired:
redis_client.expire(lock_key, expire)
logger.debug("Lock acquired: %s", lock_key)
return True
logger.debug("Lock already held: %s — skipping.", lock_key)
return False
def _release_lock(lock_key: str) -> None:
"""Release a Redis-based distributed lock."""
redis_client.delete(lock_key)
logger.debug("Lock released: %s", lock_key)
# ---------------------------------------------------------------------------
# Cache helpers
# ---------------------------------------------------------------------------
def _load_cache(cache_file: str) -> dict[str, str]:
"""Load the set of already-processed file identifiers from a JSON cache file."""
if os.path.exists(cache_file):
try:
with open(cache_file) as f:
data: dict[str, str] = json.load(f)
return _evict_old_entries(data)
except (json.JSONDecodeError, OSError):
logger.warning("Failed to read cache file %s — starting fresh.", cache_file)
return {}
def _save_cache(cache_file: str, data: dict[str, str]) -> None:
"""Persist the processed-file cache to disk."""
try:
with open(cache_file, "w") as f:
json.dump(data, f, indent=2)
except OSError as exc:
logger.error("Failed to write cache file %s: %s", cache_file, exc)
def _evict_old_entries(data: dict[str, str]) -> dict[str, str]:
"""Remove entries older than CACHE_RETENTION_DAYS to prevent unbounded growth."""
cutoff = datetime.now(timezone.utc) - timedelta(days=_CACHE_RETENTION_DAYS)
result = {}
for key, date_str in data.items():
try:
dt = datetime.fromisoformat(date_str)
if dt.tzinfo is None:
dt = dt.replace(tzinfo=timezone.utc)
if dt > cutoff:
result[key] = date_str
except (ValueError, TypeError):
pass # Skip entries with malformed timestamps
return result
def _mark_processed(cache: dict[str, str], key: str) -> None:
"""Add a file identifier to the in-memory cache dict."""
cache[key] = datetime.now(timezone.utc).isoformat()
# ---------------------------------------------------------------------------
# File type helpers
# ---------------------------------------------------------------------------
def _is_allowed_file(filename: str) -> bool:
"""Return True if the file should be ingested based on its name/extension."""
ext = os.path.splitext(filename)[1].lower()
return ext in ALLOWED_EXTENSIONS or filename.lower().endswith(".pdf")
def _enqueue_file(file_path: str, *, filename: str | None = None) -> None:
"""Enqueue a local file path for document processing."""
fname = filename or os.path.basename(file_path)
_, ext = os.path.splitext(fname)
mime_check = ext.lower() in {".pdf"}
if mime_check or fname.lower().endswith(".pdf"):
process_document.delay(file_path)
logger.info("Enqueued for processing: %s", fname)
else:
convert_to_pdf.delay(file_path)
logger.info("Enqueued for PDF conversion: %s", fname)
# ---------------------------------------------------------------------------
# Local filesystem watch folder scanning
# ---------------------------------------------------------------------------
def _scan_local_folder(folder_path: str, cache: dict[str, str], delete_after: bool) -> int:
"""
Scan a single local directory for new, allowed files.
Files already present in *cache* (keyed by their absolute path) are skipped.
Returns the number of files newly enqueued.
"""
if not os.path.isdir(folder_path):
logger.warning("Watch folder does not exist or is not a directory: %s", folder_path)
return 0
count = 0
try:
entries = os.scandir(folder_path)
except PermissionError as exc:
logger.error("Cannot scan watch folder %s: %s", folder_path, exc)
return 0
for entry in entries:
if not entry.is_file(follow_symlinks=True):
continue
if not _is_allowed_file(entry.name):
logger.debug("Skipping unsupported file type: %s", entry.name)
continue
abs_path = entry.path
if abs_path in cache:
logger.debug("Already processed: %s", abs_path)
continue
# Copy file to workdir before enqueueing so the original isn't locked
dest_filename = f"wf_{entry.name}"
dest_path = os.path.join(settings.workdir, dest_filename)
# Avoid overwriting if a file with the same name is already there
if os.path.exists(dest_path):
base, ext2 = os.path.splitext(dest_filename)
dest_path = os.path.join(settings.workdir, f"{base}_{int(datetime.now().timestamp())}{ext2}")
try:
import shutil
shutil.copy2(abs_path, dest_path)
except OSError as exc:
logger.error("Failed to copy %s to workdir: %s", abs_path, exc)
continue
_enqueue_file(dest_path)
_mark_processed(cache, abs_path)
count += 1
if delete_after:
try:
os.remove(abs_path)
logger.info("Deleted source file after ingestion: %s", abs_path)
except OSError as exc:
logger.warning("Could not delete source file %s: %s", abs_path, exc)
return count
# ---------------------------------------------------------------------------
# FTP watch folder scanning
# ---------------------------------------------------------------------------
def _connect_ftp() -> ftplib.FTP | None:
"""Establish an FTP/FTPS connection using the global FTP settings."""
host = getattr(settings, "ftp_host", None)
port = getattr(settings, "ftp_port", 21) or 21
username = getattr(settings, "ftp_username", None)
password = getattr(settings, "ftp_password", None)
if not (host and username and password):
logger.warning("FTP ingest: connection settings incomplete — skipping.")
return None
use_tls: bool = getattr(settings, "ftp_use_tls", True)
allow_plaintext: bool = getattr(settings, "ftp_allow_plaintext", True)
if use_tls:
try:
ftp = ftplib.FTP_TLS() # noqa: S321
ftp.connect(host=host, port=port)
ftp.login(user=username, passwd=password)
ftp.prot_p()
logger.debug("FTP ingest: connected via FTPS to %s:%s", host, port)
return ftp
except Exception as exc:
if not allow_plaintext:
logger.error("FTP ingest: FTPS failed and plaintext not allowed: %s", exc)
return None
logger.warning("FTP ingest: FTPS failed, falling back to plain FTP: %s", exc)
try:
ftp = ftplib.FTP() # nosec B321 # noqa: S321
ftp.connect(host=host, port=port)
ftp.login(user=username, passwd=password)
logger.debug("FTP ingest: connected via plain FTP to %s:%s", host, port)
return ftp
except Exception as exc:
logger.error("FTP ingest: connection failed: %s", exc)
return None
def _scan_ftp_folder(ftp: ftplib.FTP, remote_folder: str, cache: dict[str, str], delete_after: bool) -> int:
"""
List *remote_folder* on the FTP server and download new allowed files to workdir.
Returns the number of files newly enqueued.
"""
count = 0
try:
ftp.cwd(remote_folder)
except ftplib.error_perm as exc:
logger.error("FTP ingest: cannot CWD to %s: %s", remote_folder, exc)
return 0
try:
filenames: list[str] = ftp.nlst()
except Exception as exc:
logger.error("FTP ingest: NLST failed on %s: %s", remote_folder, exc)
return 0
for filename in filenames:
if not _is_allowed_file(filename):
logger.debug("FTP ingest: skipping %s (unsupported type)", filename)
continue
cache_key = f"ftp:{remote_folder}/{filename}"
if cache_key in cache:
logger.debug("FTP ingest: already processed %s", filename)
continue
dest_path = os.path.join(settings.workdir, f"ftp_{filename}")
if os.path.exists(dest_path):
base, ext2 = os.path.splitext(f"ftp_{filename}")
dest_path = os.path.join(settings.workdir, f"{base}_{int(datetime.now().timestamp())}{ext2}")
try:
with open(dest_path, "wb") as local_file:
ftp.retrbinary(f"RETR {filename}", local_file.write)
logger.info("FTP ingest: downloaded %s to %s", filename, dest_path)
except Exception as exc:
logger.error("FTP ingest: failed to download %s: %s", filename, exc)
if os.path.exists(dest_path):
os.remove(dest_path)
continue
_enqueue_file(dest_path, filename=filename)
_mark_processed(cache, cache_key)
count += 1
if delete_after:
try:
ftp.delete(filename)
logger.info("FTP ingest: deleted remote file %s after ingestion", filename)
except Exception as exc:
logger.warning("FTP ingest: could not delete remote file %s: %s", filename, exc)
return count
# ---------------------------------------------------------------------------
# SFTP watch folder scanning
# ---------------------------------------------------------------------------
def _get_sftp_connection():
"""
Establish an SFTP connection using the global SFTP settings.
Returns (ssh_client, sftp_client) tuple, or (None, None) on failure.
"""
import paramiko
host = getattr(settings, "sftp_host", None)
port = getattr(settings, "sftp_port", 22) or 22
username = getattr(settings, "sftp_username", None)
if not (host and username):
logger.warning("SFTP ingest: connection settings incomplete — skipping.")
return None, None
ssh = paramiko.SSHClient()
if getattr(settings, "sftp_disable_host_key_verification", False):
logger.warning(
"SFTP ingest: host key verification is DISABLED — connections are vulnerable to MITM attacks. "
"Set SFTP_DISABLE_HOST_KEY_VERIFICATION=False for production use."
)
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy()) # nosec B507 # noqa: S507
else:
ssh.load_system_host_keys()
ssh.set_missing_host_key_policy(paramiko.RejectPolicy())
connect_kwargs: dict = {"hostname": host, "port": port, "username": username}
sftp_key_path = getattr(settings, "sftp_private_key", None)
sftp_key_passphrase = getattr(settings, "sftp_private_key_passphrase", None)
sftp_password = getattr(settings, "sftp_password", None)
if sftp_key_path and os.path.exists(sftp_key_path):
connect_kwargs["key_filename"] = sftp_key_path
if sftp_key_passphrase:
connect_kwargs["passphrase"] = sftp_key_passphrase
elif sftp_password:
connect_kwargs["password"] = sftp_password
else:
logger.warning("SFTP ingest: no password or private key configured — skipping.")
return None, None
try:
ssh.connect(**connect_kwargs)
sftp = ssh.open_sftp()
logger.debug("SFTP ingest: connected to %s:%s", host, port)
return ssh, sftp
except Exception as exc:
logger.error("SFTP ingest: connection failed: %s", exc)
try:
ssh.close()
except Exception:
pass
return None, None
def _scan_sftp_folder(sftp, remote_folder: str, cache: dict[str, str], delete_after: bool) -> int:
"""
List *remote_folder* on the SFTP server and download new allowed files to workdir.
Returns the number of files newly enqueued.
"""
import stat as stat_module
count = 0
try:
entries = sftp.listdir_attr(remote_folder)
except Exception as exc:
logger.error("SFTP ingest: cannot list %s: %s", remote_folder, exc)
return 0
for attr in entries:
# Skip directories
if stat_module.S_ISDIR(attr.st_mode or 0):
continue
filename = attr.filename
if not _is_allowed_file(filename):
logger.debug("SFTP ingest: skipping %s (unsupported type)", filename)
continue
remote_path = f"{remote_folder}/{filename}".replace("//", "/")
cache_key = f"sftp:{remote_path}"
if cache_key in cache:
logger.debug("SFTP ingest: already processed %s", filename)
continue
dest_path = os.path.join(settings.workdir, f"sftp_{filename}")
if os.path.exists(dest_path):
base, ext2 = os.path.splitext(f"sftp_{filename}")
dest_path = os.path.join(settings.workdir, f"{base}_{int(datetime.now().timestamp())}{ext2}")
try:
sftp.get(remote_path, dest_path)
logger.info("SFTP ingest: downloaded %s to %s", remote_path, dest_path)
except Exception as exc:
logger.error("SFTP ingest: failed to download %s: %s", remote_path, exc)
if os.path.exists(dest_path):
os.remove(dest_path)
continue
_enqueue_file(dest_path, filename=filename)
_mark_processed(cache, cache_key)
count += 1
if delete_after:
try:
sftp.remove(remote_path)
logger.info("SFTP ingest: deleted remote file %s after ingestion", remote_path)
except Exception as exc:
logger.warning("SFTP ingest: could not delete remote file %s: %s", remote_path, exc)
return count
# ---------------------------------------------------------------------------
# Main Celery tasks
# ---------------------------------------------------------------------------
@shared_task
def scan_local_watch_folders() -> dict:
"""
Celery task: scan all configured local filesystem watch folders for new files.
Reads the WATCH_FOLDERS setting (comma-separated list of absolute directory paths)
and enqueues any new allowed files for document processing. Already-processed files
are tracked in a JSON cache file to prevent duplicate ingestion.
"""
watch_folders_raw: str | None = getattr(settings, "watch_folders", None)
if not watch_folders_raw:
logger.debug("No local watch folders configured — nothing to scan.")
return {"status": "skipped", "reason": "WATCH_FOLDERS not configured"}
folder_paths = [p.strip() for p in watch_folders_raw.split(",") if p.strip()]
if not folder_paths:
return {"status": "skipped", "reason": "WATCH_FOLDERS is empty"}
delete_after: bool = getattr(settings, "watch_folder_delete_after_process", False)
cache = _load_cache(WATCH_FOLDER_CACHE_FILE)
total = 0
for folder_path in folder_paths:
logger.info("Scanning local watch folder: %s", folder_path)
n = _scan_local_folder(folder_path, cache, delete_after)
logger.info("Local watch folder %s: %d new file(s) enqueued.", folder_path, n)
total += n
_save_cache(WATCH_FOLDER_CACHE_FILE, cache)
return {"status": "ok", "files_enqueued": total, "folders_scanned": len(folder_paths)}
@shared_task
def scan_ftp_watch_folder() -> dict:
"""
Celery task: scan the configured FTP ingest folder for new files.
Uses the existing FTP connection settings and FTP_INGEST_FOLDER to poll for
new documents to ingest. Downloaded files are enqueued for processing and
(optionally) deleted from the server after download.
"""
if not getattr(settings, "ftp_ingest_enabled", False):
logger.debug("FTP ingest is disabled — skipping.")
return {"status": "skipped", "reason": "FTP_INGEST_ENABLED is False"}
ingest_folder: str | None = getattr(settings, "ftp_ingest_folder", None)
if not ingest_folder:
logger.warning("FTP ingest enabled but FTP_INGEST_FOLDER is not set — skipping.")
return {"status": "skipped", "reason": "FTP_INGEST_FOLDER not configured"}
delete_after: bool = getattr(settings, "ftp_ingest_delete_after_process", False)
cache = _load_cache(FTP_INGEST_CACHE_FILE)
ftp = _connect_ftp()
if ftp is None:
return {"status": "error", "reason": "FTP connection failed"}
try:
n = _scan_ftp_folder(ftp, ingest_folder, cache, delete_after)
finally:
try:
ftp.quit()
except Exception:
pass
_save_cache(FTP_INGEST_CACHE_FILE, cache)
logger.info("FTP ingest: %d new file(s) enqueued from %s.", n, ingest_folder)
return {"status": "ok", "files_enqueued": n, "folder": ingest_folder}
@shared_task
def scan_sftp_watch_folder() -> dict:
"""
Celery task: scan the configured SFTP ingest folder for new files.
Uses the existing SFTP connection settings and SFTP_INGEST_FOLDER to poll for
new documents to ingest. Downloaded files are enqueued for processing and
(optionally) deleted from the server after download.
"""
if not getattr(settings, "sftp_ingest_enabled", False):
logger.debug("SFTP ingest is disabled — skipping.")
return {"status": "skipped", "reason": "SFTP_INGEST_ENABLED is False"}
ingest_folder: str | None = getattr(settings, "sftp_ingest_folder", None)
if not ingest_folder:
logger.warning("SFTP ingest enabled but SFTP_INGEST_FOLDER is not set — skipping.")
return {"status": "skipped", "reason": "SFTP_INGEST_FOLDER not configured"}
delete_after: bool = getattr(settings, "sftp_ingest_delete_after_process", False)
cache = _load_cache(SFTP_INGEST_CACHE_FILE)
ssh, sftp = _get_sftp_connection()
if sftp is None:
return {"status": "error", "reason": "SFTP connection failed"}
try:
n = _scan_sftp_folder(sftp, ingest_folder, cache, delete_after)
finally:
try:
sftp.close()
except Exception:
pass
try:
ssh.close()
except Exception:
pass
_save_cache(SFTP_INGEST_CACHE_FILE, cache)
logger.info("SFTP ingest: %d new file(s) enqueued from %s.", n, ingest_folder)
return {"status": "ok", "files_enqueued": n, "folder": ingest_folder}
@shared_task
def scan_all_watch_folders() -> dict:
"""
Main periodic Celery task that runs all watch-folder scans.
Acquires a Redis lock to prevent concurrent runs, then sequentially:
1. Scans local filesystem watch folders
2. Scans FTP ingest folder (if enabled)
3. Scans SFTP ingest folder (if enabled)
"""
if not _acquire_lock(WATCH_FOLDER_LOCK_KEY):
logger.info("Watch folder scan already running — skipping this cycle.")
return {"status": "skipped", "reason": "lock held"}
results: dict = {}
try:
results["local"] = scan_local_watch_folders()
results["ftp"] = scan_ftp_watch_folder()
results["sftp"] = scan_sftp_watch_folder()
finally:
_release_lock(WATCH_FOLDER_LOCK_KEY)
return {"status": "ok", "results": results}