feat: Brief附件/项目平台/规则AI解析/消息中心修复 + 项目创建通知
- Brief 支持代理商附件上传 (迁移 007) - 项目新增 platform 字段 (迁移 008)，前端创建/展示平台信息 - 修复 AI 规则解析：处理中文引号导致 JSON 解析失败的问题 - 修复消息中心崩溃：补全后端消息类型映射 + fallback 保护 - 项目创建时自动发送消息通知 - .gitignore 排除 backend/data/ 数据库文件 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Your Name
co-authored by
Claude Opus 4.6
Claude Opus 4.6
parent
58aed5f201
commit
4c9b2f1263
@@ -48,7 +48,7 @@ class OpenAICompatibleClient:
|
||||
base_url: str,
|
||||
api_key: str,
|
||||
provider: str = "openai",
|
||||
timeout: float = 60.0,
|
||||
timeout: float = 180.0,
|
||||
):
|
||||
self.base_url = base_url.rstrip("/")
|
||||
self.api_key = api_key
|
||||
|
||||
@@ -17,6 +17,9 @@ class DocumentParser:
|
||||
"""
|
||||
下载文档并解析为纯文本
|
||||
|
||||
优先使用 TOS SDK 直接下载（私有桶无需签名），
|
||||
回退到 HTTP 预签名 URL 下载。
|
||||
|
||||
Args:
|
||||
document_url: 文档 URL (TOS)
|
||||
document_name: 原始文件名（用于判断格式）
|
||||
@@ -24,16 +27,19 @@ class DocumentParser:
|
||||
Returns:
|
||||
提取的纯文本
|
||||
"""
|
||||
# 下载到临时文件
|
||||
tmp_path: Optional[str] = None
|
||||
try:
|
||||
async with httpx.AsyncClient(timeout=60.0) as client:
|
||||
resp = await client.get(document_url)
|
||||
resp.raise_for_status()
|
||||
|
||||
ext = document_name.rsplit(".", 1)[-1].lower() if "." in document_name else ""
|
||||
|
||||
# 优先用 TOS SDK 直接下载（后端有 AK/SK，无需签名 URL）
|
||||
content = await DocumentParser._download_via_tos_sdk(document_url)
|
||||
|
||||
if content is None:
|
||||
# 回退：生成预签名 URL 后用 HTTP 下载
|
||||
content = await DocumentParser._download_via_signed_url(document_url)
|
||||
|
||||
with tempfile.NamedTemporaryFile(delete=False, suffix=f".{ext}") as tmp:
|
||||
tmp.write(resp.content)
|
||||
tmp.write(content)
|
||||
tmp_path = tmp.name
|
||||
|
||||
return DocumentParser.parse_file(tmp_path, document_name)
|
||||
@@ -41,6 +47,75 @@ class DocumentParser:
|
||||
if tmp_path and os.path.exists(tmp_path):
|
||||
os.unlink(tmp_path)
|
||||
|
||||
@staticmethod
|
||||
async def download_and_get_images(document_url: str, document_name: str) -> Optional[list[str]]:
|
||||
"""
|
||||
下载 PDF 并将页面转为 base64 图片列表（用于图片型 PDF 的 AI 视觉解析）。
|
||||
非 PDF 或非图片型 PDF 返回 None。
|
||||
"""
|
||||
ext = document_name.rsplit(".", 1)[-1].lower() if "." in document_name else ""
|
||||
if ext != "pdf":
|
||||
return None
|
||||
|
||||
tmp_path: Optional[str] = None
|
||||
try:
|
||||
file_content = await DocumentParser._download_via_tos_sdk(document_url)
|
||||
if file_content is None:
|
||||
file_content = await DocumentParser._download_via_signed_url(document_url)
|
||||
|
||||
with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp:
|
||||
tmp.write(file_content)
|
||||
tmp_path = tmp.name
|
||||
|
||||
if DocumentParser.is_image_pdf(tmp_path):
|
||||
return DocumentParser.pdf_to_images_base64(tmp_path)
|
||||
return None
|
||||
finally:
|
||||
if tmp_path and os.path.exists(tmp_path):
|
||||
os.unlink(tmp_path)
|
||||
|
||||
@staticmethod
|
||||
async def _download_via_tos_sdk(document_url: str) -> Optional[bytes]:
|
||||
"""通过 TOS SDK 直接下载文件（私有桶安全访问）"""
|
||||
try:
|
||||
from app.config import settings
|
||||
from app.services.oss import parse_file_key_from_url
|
||||
import tos as tos_sdk
|
||||
|
||||
if not settings.TOS_ACCESS_KEY_ID or not settings.TOS_SECRET_ACCESS_KEY:
|
||||
return None
|
||||
|
||||
file_key = parse_file_key_from_url(document_url)
|
||||
if not file_key or file_key == document_url:
|
||||
return None
|
||||
|
||||
region = settings.TOS_REGION
|
||||
endpoint = settings.TOS_ENDPOINT or f"tos-cn-{region}.volces.com"
|
||||
|
||||
client = tos_sdk.TosClientV2(
|
||||
ak=settings.TOS_ACCESS_KEY_ID,
|
||||
sk=settings.TOS_SECRET_ACCESS_KEY,
|
||||
endpoint=f"https://{endpoint}",
|
||||
region=region,
|
||||
)
|
||||
resp = client.get_object(bucket=settings.TOS_BUCKET_NAME, key=file_key)
|
||||
return resp.read()
|
||||
except Exception:
|
||||
return None
|
||||
|
||||
@staticmethod
|
||||
async def _download_via_signed_url(document_url: str) -> bytes:
|
||||
"""生成预签名 URL 后通过 HTTP 下载"""
|
||||
from app.services.oss import generate_presigned_url, parse_file_key_from_url
|
||||
|
||||
file_key = parse_file_key_from_url(document_url)
|
||||
signed_url = generate_presigned_url(file_key, expire_seconds=300)
|
||||
|
||||
async with httpx.AsyncClient(timeout=60.0) as client:
|
||||
resp = await client.get(signed_url)
|
||||
resp.raise_for_status()
|
||||
return resp.content
|
||||
|
||||
@staticmethod
|
||||
def parse_file(file_path: str, file_name: str) -> str:
|
||||
"""
|
||||
@@ -68,16 +143,73 @@ class DocumentParser:
|
||||
|
||||
@staticmethod
|
||||
def _parse_pdf(path: str) -> str:
|
||||
"""pdfplumber 提取 PDF 文本"""
|
||||
import pdfplumber
|
||||
"""PyMuPDF 提取 PDF 文本，回退 pdfplumber"""
|
||||
import fitz
|
||||
|
||||
texts = []
|
||||
with pdfplumber.open(path) as pdf:
|
||||
for page in pdf.pages:
|
||||
text = page.extract_text()
|
||||
if text:
|
||||
texts.append(text)
|
||||
return "\n".join(texts)
|
||||
doc = fitz.open(path)
|
||||
for page in doc:
|
||||
text = page.get_text()
|
||||
if text and text.strip():
|
||||
texts.append(text.strip())
|
||||
doc.close()
|
||||
|
||||
result = "\n".join(texts)
|
||||
|
||||
# 如果 PyMuPDF 提取文本太少，回退 pdfplumber
|
||||
if len(result.strip()) < 100:
|
||||
try:
|
||||
import pdfplumber
|
||||
texts2 = []
|
||||
with pdfplumber.open(path) as pdf:
|
||||
for page in pdf.pages:
|
||||
text = page.extract_text()
|
||||
if text:
|
||||
texts2.append(text)
|
||||
fallback = "\n".join(texts2)
|
||||
if len(fallback.strip()) > len(result.strip()):
|
||||
result = fallback
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
return result
|
||||
|
||||
@staticmethod
|
||||
def pdf_to_images_base64(path: str, max_pages: int = 5, dpi: int = 150) -> list[str]:
|
||||
"""
|
||||
将 PDF 页面渲染为图片并返回 base64 编码列表。
|
||||
用于处理扫描件/图片型 PDF。
|
||||
"""
|
||||
import fitz
|
||||
import base64
|
||||
|
||||
images = []
|
||||
doc = fitz.open(path)
|
||||
for i, page in enumerate(doc):
|
||||
if i >= max_pages:
|
||||
break
|
||||
zoom = dpi / 72
|
||||
mat = fitz.Matrix(zoom, zoom)
|
||||
pix = page.get_pixmap(matrix=mat)
|
||||
img_bytes = pix.tobytes("png")
|
||||
b64 = base64.b64encode(img_bytes).decode()
|
||||
images.append(b64)
|
||||
doc.close()
|
||||
return images
|
||||
|
||||
@staticmethod
|
||||
def is_image_pdf(path: str) -> bool:
|
||||
"""判断 PDF 是否为扫描件/图片型（文本内容极少）"""
|
||||
import fitz
|
||||
|
||||
doc = fitz.open(path)
|
||||
total_text = ""
|
||||
for page in doc:
|
||||
total_text += page.get_text()
|
||||
doc.close()
|
||||
# 去掉页码等噪音后，有效文字少于 200 字符视为图片 PDF
|
||||
cleaned = "".join(c for c in total_text if c.strip())
|
||||
return len(cleaned) < 200
|
||||
|
||||
@staticmethod
|
||||
def _parse_docx(path: str) -> str:
|
||||
|
||||
Reference in New Issue
Block a user