Files
WeKnora/docreader/main.py
2025-11-05 12:07:39 +08:00

404 lines
15 KiB
Python

import os
import sys
import logging
from concurrent import futures
import traceback
import grpc
import uuid
import atexit
from grpc_health.v1 import health_pb2_grpc
from grpc_health.v1.health import HealthServicer
# Add parent directory to Python path
current_dir = os.path.dirname(os.path.abspath(__file__))
parent_dir = os.path.dirname(current_dir)
if parent_dir not in sys.path:
sys.path.insert(0, parent_dir)
from proto.docreader_pb2 import ReadResponse, Chunk, Image
from proto import docreader_pb2_grpc
from parser import Parser, OCREngine
from parser.config import ChunkingConfig
from utils.request import request_id_context, init_logging_request_id
# --- Encoding utilities: sanitize strings to valid UTF-8 and (optionally) multi-encoding read ---
import re
from typing import Optional
try:
# Optional dependency for charset detection; install via `pip install charset-normalizer`
from charset_normalizer import from_bytes as _cn_from_bytes # type: ignore
except Exception: # pragma: no cover
_cn_from_bytes = None # type: ignore
# Surrogate range U+D800..U+DFFF are invalid Unicode scalar values and cannot be encoded to UTF-8
_SURROGATE_RE = re.compile(r"[\ud800-\udfff]")
def to_valid_utf8_text(s: Optional[str]) -> str:
"""Return a UTF-8 safe string for protobuf.
- Replace any surrogate code points with U+FFFD
- Re-encode with errors='replace' to ensure valid UTF-8
"""
if not s:
return ""
s = _SURROGATE_RE.sub("\ufffd", s)
return s.encode("utf-8", errors="replace").decode("utf-8")
def read_text_with_fallback(file_path: str) -> str:
"""Read text from file supporting multiple encodings with graceful fallback.
This server currently receives bytes over gRPC and delegates decoding to the parser.
This helper is provided for future local-file reads if needed.
"""
with open(file_path, "rb") as f:
raw = f.read()
if _cn_from_bytes is not None:
try:
result = _cn_from_bytes(raw).best()
if result:
return str(result)
except Exception:
pass
for enc in ("utf-8", "gb18030", "latin-1"):
try:
return raw.decode(enc, errors="replace")
except UnicodeDecodeError:
continue
return raw.decode("utf-8", errors="replace")
# Ensure no existing handlers
for handler in logging.root.handlers[:]:
logging.root.removeHandler(handler)
# Configure logging - use stdout
handler = logging.StreamHandler(sys.stdout)
logging.root.addHandler(handler)
logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG)
logger.info("Initializing server logging")
# Initialize request ID logging
init_logging_request_id()
# Set max message size to 50MB
MAX_MESSAGE_LENGTH = 50 * 1024 * 1024
parser = Parser()
class DocReaderServicer(docreader_pb2_grpc.DocReaderServicer):
def __init__(self):
super().__init__()
self.parser = Parser()
def ReadFromFile(self, request, context):
# Get or generate request ID
request_id = (
request.request_id
if hasattr(request, "request_id") and request.request_id
else str(uuid.uuid4())
)
# Use request ID context
with request_id_context(request_id):
try:
# Get file type
file_type = (
request.file_type or os.path.splitext(request.file_name)[1][1:]
)
logger.info(
f"Received ReadFromFile request for file: {request.file_name}, type: {file_type}"
)
logger.info(f"File content size: {len(request.file_content)} bytes")
# Create chunking config
chunk_size = request.read_config.chunk_size or 512
chunk_overlap = request.read_config.chunk_overlap or 50
separators = request.read_config.separators or ["\n\n", "\n", ""]
enable_multimodal = request.read_config.enable_multimodal or False
logger.info(
f"Using chunking config: size={chunk_size}, overlap={chunk_overlap}, "
f"multimodal={enable_multimodal}"
)
# Get Storage and VLM config from request
storage_config = None
vlm_config = None
sc = request.read_config.storage_config
# Keep parser-side key name as cos_config for backward compatibility
storage_config = {
"provider": "minio" if sc.provider == 2 else "cos",
"region": sc.region,
"bucket_name": sc.bucket_name,
"access_key_id": sc.access_key_id,
"secret_access_key": sc.secret_access_key,
"app_id": sc.app_id,
"path_prefix": sc.path_prefix,
}
logger.info(
f"Using Storage config: provider={storage_config.get('provider')}, bucket={storage_config['bucket_name']}"
)
vlm_config = {
"model_name": request.read_config.vlm_config.model_name,
"base_url": request.read_config.vlm_config.base_url,
"api_key": request.read_config.vlm_config.api_key or "",
"interface_type": request.read_config.vlm_config.interface_type
or "openai",
}
logger.info(
f"Using VLM config: model={vlm_config['model_name']}, "
f"base_url={vlm_config['base_url']}, "
f"interface_type={vlm_config['interface_type']}"
)
chunking_config = ChunkingConfig(
chunk_size=chunk_size,
chunk_overlap=chunk_overlap,
separators=separators,
enable_multimodal=enable_multimodal,
storage_config=storage_config,
vlm_config=vlm_config,
)
# Parse file
logger.info(f"Starting file parsing process")
result = self.parser.parse_file(
request.file_name, file_type, request.file_content, chunking_config
)
if not result:
error_msg = "Failed to parse file"
logger.error(error_msg)
context.set_code(grpc.StatusCode.INTERNAL)
context.set_details(error_msg)
return ReadResponse()
# Convert to protobuf message
logger.info(
f"Successfully parsed file {request.file_name}, returning {len(result.chunks)} chunks"
)
# Build response, including image info
response = ReadResponse(
chunks=[
self._convert_chunk_to_proto(chunk) for chunk in result.chunks
]
)
logger.info(f"Response size: {response.ByteSize()} bytes")
return response
except Exception as e:
error_msg = f"Error reading file: {str(e)}"
logger.error(error_msg)
logger.info(f"Detailed traceback: {traceback.format_exc()}")
context.set_code(grpc.StatusCode.INTERNAL)
context.set_details(str(e))
return ReadResponse(error=str(e))
def ReadFromURL(self, request, context):
# Get or generate request ID
request_id = (
request.request_id
if hasattr(request, "request_id") and request.request_id
else str(uuid.uuid4())
)
# Use request ID context
with request_id_context(request_id):
try:
logger.info(f"Received ReadFromURL request for URL: {request.url}")
# Create chunking config
chunk_size = request.read_config.chunk_size or 512
chunk_overlap = request.read_config.chunk_overlap or 50
separators = request.read_config.separators or ["\n\n", "\n", ""]
enable_multimodal = request.read_config.enable_multimodal or False
logger.info(
f"Using chunking config: size={chunk_size}, overlap={chunk_overlap}, "
f"multimodal={enable_multimodal}"
)
# Get Storage and VLM config from request
storage_config = None
vlm_config = None
sc = request.read_config.storage_config
storage_config = {
"provider": "minio" if sc.provider == 2 else "cos",
"region": sc.region,
"bucket_name": sc.bucket_name,
"access_key_id": sc.access_key_id,
"secret_access_key": sc.secret_access_key,
"app_id": sc.app_id,
"path_prefix": sc.path_prefix,
}
logger.info(
f"Using Storage config: provider={storage_config.get('provider')}, bucket={storage_config['bucket_name']}"
)
vlm_config = {
"model_name": request.read_config.vlm_config.model_name,
"base_url": request.read_config.vlm_config.base_url,
"api_key": request.read_config.vlm_config.api_key or "",
"interface_type": request.read_config.vlm_config.interface_type
or "openai",
}
logger.info(
f"Using VLM config: model={vlm_config['model_name']}, "
f"base_url={vlm_config['base_url']}, "
f"interface_type={vlm_config['interface_type']}"
)
chunking_config = ChunkingConfig(
chunk_size=chunk_size,
chunk_overlap=chunk_overlap,
separators=separators,
enable_multimodal=enable_multimodal,
storage_config=storage_config,
vlm_config=vlm_config,
)
# Parse URL
logger.info(f"Starting URL parsing process")
result = self.parser.parse_url(
request.url, request.title, chunking_config
)
if not result:
error_msg = "Failed to parse URL"
logger.error(error_msg)
context.set_code(grpc.StatusCode.INTERNAL)
context.set_details(error_msg)
return ReadResponse(error=error_msg)
# Convert to protobuf message, including image info
logger.info(
f"Successfully parsed URL {request.url}, returning {len(result.chunks)} chunks"
)
response = ReadResponse(
chunks=[
self._convert_chunk_to_proto(chunk) for chunk in result.chunks
]
)
logger.info(f"Response size: {response.ByteSize()} bytes")
return response
except Exception as e:
error_msg = f"Error reading URL: {str(e)}"
logger.error(error_msg)
logger.info(f"Detailed traceback: {traceback.format_exc()}")
context.set_code(grpc.StatusCode.INTERNAL)
context.set_details(str(e))
return ReadResponse(error=str(e))
def _convert_chunk_to_proto(self, chunk):
"""Convert internal Chunk object to protobuf Chunk message
Ensures all string fields are valid UTF-8 for protobuf (no lone surrogates).
"""
# Clean helper for strings
_c = to_valid_utf8_text
proto_chunk = Chunk(
content=_c(getattr(chunk, "content", None)),
seq=getattr(chunk, "seq", 0),
start=getattr(chunk, "start", 0),
end=getattr(chunk, "end", 0),
)
# If chunk has images attribute and is not empty, add image info
if hasattr(chunk, "images") and chunk.images:
logger.info(
f"Adding {len(chunk.images)} images to chunk {getattr(chunk, 'seq', 0)}"
)
for img_info in chunk.images:
# img_info expected as dict
proto_image = Image(
url=_c(img_info.get("cos_url", "")),
caption=_c(img_info.get("caption", "")),
ocr_text=_c(img_info.get("ocr_text", "")),
original_url=_c(img_info.get("original_url", "")),
start=int(img_info.get("start", 0) or 0),
end=int(img_info.get("end", 0) or 0),
)
proto_chunk.images.append(proto_image)
return proto_chunk
def init_ocr_engine(ocr_backend, ocr_config):
"""Initialize OCR engine"""
try:
logger.info(f"Initializing OCR engine with backend: {ocr_backend}")
ocr_engine = OCREngine.get_instance(backend_type=ocr_backend, **ocr_config)
if ocr_engine:
logger.info("OCR engine initialized successfully")
return True
else:
logger.error("OCR engine initialization failed")
return False
except Exception as e:
logger.error(f"Error initializing OCR engine: {str(e)}")
return False
def main():
init_ocr_engine(
os.getenv("OCR_BACKEND", "paddle"),
{
"OCR_API_BASE_URL": os.getenv("OCR_API_BASE_URL", ""),
},
)
# Set max number of worker threads
max_workers = int(os.environ.get("GRPC_MAX_WORKERS", "4"))
logger.info(f"Starting DocReader service with {max_workers} worker threads")
# Get port number
port = os.environ.get("GRPC_PORT", "50051")
# Create server
server = grpc.server(
futures.ThreadPoolExecutor(max_workers=max_workers),
options=[
("grpc.max_send_message_length", MAX_MESSAGE_LENGTH),
("grpc.max_receive_message_length", MAX_MESSAGE_LENGTH),
],
)
# Register services
docreader_pb2_grpc.add_DocReaderServicer_to_server(DocReaderServicer(), server)
# Register health check service
health_servicer = HealthServicer()
health_pb2_grpc.add_HealthServicer_to_server(health_servicer, server)
# Set listen address
server.add_insecure_port(f"[::]:{port}")
# Start service
server.start()
logger.info(f"Server started on port {port}")
logger.info("Server is ready to accept connections")
try:
# Wait for service termination
server.wait_for_termination()
except KeyboardInterrupt:
logger.info("Received termination signal, shutting down server")
server.stop(0)
if __name__ == "__main__":
main()