From fd400f2be18defcf90847b32eb86b4497dbf071c Mon Sep 17 00:00:00 2001 From: shaohuzhang1 <80892890+shaohuzhang1@users.noreply.github.com> Date: Fri, 12 Dec 2025 11:57:38 +0800 Subject: [PATCH] fix: The image uploaded from the workflow knowledge base zip file cannot be parsed (#4503) --- .../impl/base_document_extract_node.py | 3 +- .../handle/impl/text/zip_split_handle.py | 128 ++++++++---------- 2 files changed, 58 insertions(+), 73 deletions(-) diff --git a/apps/application/flow/step_node/document_extract_node/impl/base_document_extract_node.py b/apps/application/flow/step_node/document_extract_node/impl/base_document_extract_node.py index b3a317724..7d196134b 100644 --- a/apps/application/flow/step_node/document_extract_node/impl/base_document_extract_node.py +++ b/apps/application/flow/step_node/document_extract_node/impl/base_document_extract_node.py @@ -53,7 +53,8 @@ class BaseDocumentExtractNode(IDocumentExtractNode): source_id=meta['application_id'] if meta['application_id'] else meta['knowledge_id'], meta=meta ) - new_file.save(file_bytes) + if not QuerySet(File).filter(id=new_file.id).exists(): + new_file.save(file_bytes) document_list = [] for doc in document: diff --git a/apps/common/handle/impl/text/zip_split_handle.py b/apps/common/handle/impl/text/zip_split_handle.py index f41d4c302..46fdea58c 100644 --- a/apps/common/handle/impl/text/zip_split_handle.py +++ b/apps/common/handle/impl/text/zip_split_handle.py @@ -101,6 +101,37 @@ def get_image_list(result_list: list, zip_files: List[str]): return image_file_list +def get_image_list_by_content(name: str, content: str, zip_files: List[str]): + image_file_list = [] + image_list = parse_md_image(content) + for image in image_list: + search = re.search("\(.*\)", image) + if search: + new_image_id = str(uuid.uuid7()) + source_image_path = search.group().replace('(', '').replace(')', '') + source_image_path = source_image_path.strip().split(" ")[0] + image_path = urljoin(name, '.' + source_image_path if source_image_path.startswith( + '/') else source_image_path) + if not zip_files.__contains__(image_path): + continue + if image_path.startswith('oss/file/') or image_path.startswith('oss/image/'): + image_id = image_path.replace('oss/file/', '').replace('oss/file/', '') + if is_valid_uuid(image_id): + image_file_list.append({'source_file': image_path, + 'image_id': image_id}) + else: + image_file_list.append({'source_file': image_path, + 'image_id': new_image_id}) + content = content.replace(source_image_path, f'./oss/file/{new_image_id}') + + else: + image_file_list.append({'source_file': image_path, + 'image_id': new_image_id}) + content = content.replace(source_image_path, f'./oss/file/{new_image_id}') + + return image_file_list, content + + def get_file_name(file_name): try: file_name_code = file_name.encode('cp437') @@ -171,17 +202,12 @@ class ZipSplitHandle(BaseSplitHandle): """ buffer = file.read() if hasattr(file, 'read') else None bytes_io = io.BytesIO(buffer) if buffer is not None else io.BytesIO(file) - md_items = [] # 存储 (md_text, source_file_path) - image_mode_list = [] - - import posixpath - - def is_image_name(name: str): - ext = posixpath.splitext(name.lower())[1] - return ext in ('.png', '.jpg', '.jpeg', '.gif', '.bmp', '.webp', '.svg') + image_list = [] + content_parts = [] with zipfile.ZipFile(bytes_io, 'r') as zip_ref: files = zip_ref.namelist() + file_content_list = [] for inner_name in files: if inner_name.endswith('/') or inner_name.startswith('__MACOSX'): continue @@ -190,77 +216,35 @@ class ZipSplitHandle(BaseSplitHandle): real_name = get_file_name(zf.name) except Exception: real_name = zf.name - raw = zf.read() - # 图片直接收集 - if is_image_name(real_name): - image_id = str(uuid.uuid7()) - fmodel = File( - id=image_id, - file_name=os.path.basename(real_name), - meta={'debug': False, 'content': raw} - ) - image_mode_list.append(fmodel) - continue # 为 split_handle 提供可重复读取的 file-like 对象 - inner_file = io.BytesIO(raw) - inner_file.name = real_name - - # 尝试使用已注册的 split handle 的 get_content - md_text = None + zf.name = real_name for split_handle in split_handles: # 准备一个简单的 get_buffer 回调，返回当前 raw - get_buffer = lambda f, _raw=raw: _raw - if split_handle.support(inner_file, get_buffer): - inner_file.seek(0) - md_text = split_handle.get_content(inner_file, save_image) + get_buffer = FileBufferHandle().get_buffer + if split_handle.support(zf, get_buffer): + row = get_buffer(zf) + md_text = split_handle.get_content(io.BytesIO(row), save_image) + file_content_list.append({'content': md_text, 'name': real_name}) break - - # 如果没有任何 split_handle 处理，按文本解码作为后备 - if md_text is None: - enc = detect(raw).get('encoding') or 'utf-8' - try: - md_text = raw.decode(enc, errors='ignore') - except Exception: - md_text = raw.decode('utf-8', errors='ignore') - - if isinstance(md_text, str) and md_text.strip(): - # 保存 md 文本与其所在的文件路径，后面统一做图片路径替换 - md_items.append((md_text, real_name)) + for file_content in file_content_list: + _image_list, content = get_image_list_by_content(file_content.get('name'), file_content.get("content"), + files) + content_parts.append(content) + for image in _image_list: + image_list.append(image) # 将收集到的图片通过回调保存（一次性） - if image_mode_list: + if image_list: + image_mode_list = [] + for image in image_list: + with zip_ref.open(image.get('source_file')) as f: + i = File( + id=image.get('image_id'), + file_name=os.path.basename(image.get('source_file')), + meta={'debug': False, 'content': f.read()} # 这里的content是二进制数据 + ) + image_mode_list.append(i) save_image(image_mode_list) - # 后处理：在每个 md 片段中将相对/绝对引用替换为已保存图片的 oss 路径 - content_parts = [] - for md_text, base_name in md_items: - image_refs = parse_md_image(md_text) - for image in image_refs: - search = re.search(r"\(.*\)", image) - if not search: - continue - source_image_path = search.group().strip("()").split(" ")[0] - - # 规范化 zip 内部路径：若以 '/' 开头，视为相对于 zip 根，否则相对于 base_name 的目录 - if source_image_path.startswith('/'): - joined = posixpath.normpath(source_image_path.lstrip('/')) - else: - base_dir = posixpath.dirname(base_name) - joined = posixpath.normpath(posixpath.join(base_dir, source_image_path)) - - # 匹配已收集图片：以文件名做匹配（zip 中的文件名通常是不含反斜杠的 POSIX 风格） - matched = None - for img_model in image_mode_list: - if img_model.file_name == posixpath.basename(joined): - matched = img_model - break - - if matched: - md_text = md_text.replace(source_image_path, f'./oss/file/{matched.id}') - - content_parts.append(md_text) - return '\n\n'.join(content_parts) - -