diff --git a/apps/admin/config_service.py b/apps/admin/config_service.py index 6034247..efc8127 100644 --- a/apps/admin/config_service.py +++ b/apps/admin/config_service.py @@ -40,6 +40,7 @@ class ConfigService: "max_save_seconds", "onedrive_proxy", "open_upload", + "mcp_require_token", "port", "s3_proxy", "server_port", diff --git a/apps/base/mcp.py b/apps/base/mcp.py new file mode 100644 index 0000000..bb5fb50 --- /dev/null +++ b/apps/base/mcp.py @@ -0,0 +1,302 @@ +"""Stateless MCP Streamable HTTP endpoint for quick share uploads.""" +import base64 +import binascii +import html +import io +import json +from html.parser import HTMLParser +from urllib.parse import quote, urlsplit + +from fastapi import APIRouter, HTTPException, Request, UploadFile +from starlette.datastructures import Headers +from starlette.responses import JSONResponse, Response + +from apps.base.auth import _require_admin_payload +from apps.base.file_validation import validate_file_type, validate_upload_file +from apps.base.services import FileUploadService, validate_file_size +from apps.base.utils import ip_limit, validate_expire_style +from core.settings import settings +from core.utils import sanitize_filename + +router = APIRouter(tags=["MCP"]) + +MCP_PROTOCOL_VERSION = "2025-03-26" +TEXT_SHARE_MAX_BYTES = 222 * 1024 +MAX_MCP_BODY_BYTES = 64 * 1024 * 1024 + + +def _jsonrpc_error(request_id, code: int, message: str) -> dict: + return {"jsonrpc": "2.0", "id": request_id, "error": {"code": code, "message": message}} + + +def _jsonrpc_result(request_id, result: dict) -> dict: + return {"jsonrpc": "2.0", "id": request_id, "result": result} + + +def _validate_origin(request: Request) -> None: + """Reject browser cross-site POSTs before parsing any upload tool.""" + origin = request.headers.get("origin") + fetch_site = request.headers.get("sec-fetch-site", "").lower() + if fetch_site == "cross-site": + raise HTTPException(status_code=403, detail="跨站请求已拒绝") + if not origin: + return # Native MCP clients do not send Origin and use no ambient cookies. + try: + parsed = urlsplit(origin) + host = request.headers.get("host", "").lower() + if ( + parsed.scheme not in {"http", "https"} + or not parsed.netloc + or parsed.netloc.lower() != host + or parsed.username is not None + or parsed.password is not None + or parsed.path not in {"", "/"} + or parsed.query + or parsed.fragment + ): + raise ValueError("origin mismatch") + except (ValueError, TypeError): + raise HTTPException(status_code=403, detail="跨站请求已拒绝") + + +async def _read_limited_body(request: Request) -> bytes: + max_size = min( + MAX_MCP_BODY_BYTES, + max(1024 * 1024, int(settings.upload_size * 4 / 3) + 256 * 1024), + ) + declared_size = request.headers.get("content-length") + if declared_size: + try: + if int(declared_size) > max_size: + raise HTTPException(status_code=413, detail="MCP 请求体过大") + except ValueError: + raise HTTPException(status_code=400, detail="Content-Length 无效") + body = bytearray() + async for chunk in request.stream(): + body.extend(chunk) + if len(body) > max_size: + raise HTTPException(status_code=413, detail="MCP 请求体过大") + return bytes(body) + + +def _tool_definitions() -> list[dict]: + return [ + { + "name": "share_text", + "description": "将文本快速上传到 FileCodeBox 并创建分享链接。", + "inputSchema": { + "type": "object", + "properties": { + "text": {"type": "string", "description": "要分享的文本"}, + "expire_value": {"type": "integer", "minimum": 1, "default": 1}, + "expire_style": {"type": "string", "enum": ["day", "hour", "minute", "forever", "count"], "default": "day"}, + }, + "required": ["text"], + }, + }, + { + "name": "share_file", + "description": "将 Base64 编码的文件上传到 FileCodeBox 并创建分享链接。", + "inputSchema": { + "type": "object", + "properties": { + "filename": {"type": "string", "description": "文件名"}, + "content_base64": {"type": "string", "description": "文件内容的标准 Base64 编码"}, + "content_type": {"type": "string", "default": "application/octet-stream"}, + "expire_value": {"type": "integer", "minimum": 1, "default": 1}, + "expire_style": {"type": "string", "enum": ["day", "hour", "minute", "forever", "count"], "default": "day"}, + }, + "required": ["filename", "content_base64"], + }, + }, + ] + + +def _share_result(request: Request, code: str, name: str) -> dict: + base_url = str(request.base_url).rstrip("/") + share_url = f"{base_url}/#/?code={quote(code, safe='')}" + markdown_name = html.escape(name).replace("[", "\\[").replace("]", "\\]") + return { + "code": code, + "name": name, + "url": share_url, + "markdown": f"[{markdown_name}]({share_url})", + } + + +async def _call_tool(request: Request, tool_name: str, arguments: dict) -> dict: + authorization = request.headers.get("authorization", "") + if settings.mcp_require_token or not settings.open_upload: + _require_admin_payload(authorization) + ip = ip_limit["upload"](request) + + if tool_name == "share_text": + text = arguments.get("text") + if not isinstance(text, str): + raise ValueError("text 必须是字符串") + if len(text.encode("utf-8")) > TEXT_SHARE_MAX_BYTES: + raise ValueError("内容过多,请改用文件分享") + expire_value = int(arguments.get("expire_value", 1)) + expire_style = str(arguments.get("expire_style", "day")) + if expire_value < 1: + raise ValueError("expire_value 必须大于 0") + validate_expire_style(expire_style) + code = await FileUploadService.create_text_share(text, expire_value, expire_style) + ip_limit["upload"].add_ip(ip) + return _share_result(request, code, "Text") + + if tool_name == "share_file": + filename = arguments.get("filename") + encoded_content = arguments.get("content_base64") + content_type = str(arguments.get("content_type", "application/octet-stream")) + if not isinstance(filename, str) or not filename.strip(): + raise ValueError("filename 不能为空") + if not isinstance(encoded_content, str): + raise ValueError("content_base64 必须是字符串") + try: + content = base64.b64decode(encoded_content, validate=True) + except (ValueError, binascii.Error): + raise ValueError("content_base64 不是有效的标准 Base64") + safe_name = await sanitize_filename(filename.rsplit("/", 1)[-1].rsplit("\\", 1)[-1]) + validate_file_type(safe_name, content_type) + upload = UploadFile(filename=safe_name, file=io.BytesIO(content), size=len(content), headers=Headers({"content-type": content_type})) + size = await validate_file_size(upload, settings.upload_size) + await validate_upload_file(upload) + expire_value = int(arguments.get("expire_value", 1)) + expire_style = str(arguments.get("expire_style", "day")) + if expire_value < 1: + raise ValueError("expire_value 必须大于 0") + validate_expire_style(expire_style) + share = await FileUploadService.create_file_share(upload, size=size, expire_value=expire_value, expire_style=expire_style) + ip_limit["upload"].add_ip(ip) + return _share_result(request, share["code"], share["name"]) + + raise LookupError(f"未知工具:{tool_name}") + + +@router.post("/mcp", include_in_schema=False) +async def mcp_endpoint(request: Request): + _validate_origin(request) + media_type = request.headers.get("content-type", "").split(";", 1)[0].strip().lower() + if media_type != "application/json": + raise HTTPException(status_code=415, detail="MCP 请求必须使用 application/json") + try: + message = json.loads(await _read_limited_body(request)) + except (json.JSONDecodeError, UnicodeDecodeError): + return JSONResponse(_jsonrpc_error(None, -32700, "Parse error"), status_code=400) + if not isinstance(message, dict) or message.get("jsonrpc") != "2.0" or not isinstance(message.get("method"), str): + return JSONResponse(_jsonrpc_error(message.get("id") if isinstance(message, dict) else None, -32600, "Invalid Request"), status_code=400) + + request_id = message.get("id") + method = message["method"] + if request_id is None and method.startswith("notifications/"): + return Response(status_code=202) + try: + if method == "initialize": + result = { + "protocolVersion": MCP_PROTOCOL_VERSION, + "capabilities": {"tools": {"listChanged": False}}, + "serverInfo": {"name": "filecodebox", "version": "1.0.0"}, + } + elif method == "ping": + result = {} + elif method == "tools/list": + result = {"tools": _tool_definitions()} + elif method == "tools/call": + params = message.get("params") + if not isinstance(params, dict) or not isinstance(params.get("name"), str): + return JSONResponse(_jsonrpc_error(request_id, -32602, "Invalid params")) + arguments = params.get("arguments", {}) + if not isinstance(arguments, dict): + return JSONResponse(_jsonrpc_error(request_id, -32602, "Invalid params")) + try: + tool_value = await _call_tool(request, params["name"], arguments) + result = {"content": [{"type": "text", "text": json.dumps(tool_value, ensure_ascii=False)}], "structuredContent": tool_value} + except LookupError as exc: + return JSONResponse(_jsonrpc_error(request_id, -32602, str(exc))) + except HTTPException as exc: + if exc.status_code in {401, 403}: + raise + result = {"content": [{"type": "text", "text": str(exc.detail)}], "isError": True} + except (TypeError, ValueError) as exc: + result = {"content": [{"type": "text", "text": str(exc)}], "isError": True} + else: + return JSONResponse(_jsonrpc_error(request_id, -32601, "Method not found")) + return JSONResponse(_jsonrpc_result(request_id, result), headers={"MCP-Protocol-Version": MCP_PROTOCOL_VERSION}) + except HTTPException as exc: + return JSONResponse( + _jsonrpc_error(request_id, -32001 if exc.status_code == 401 else -32003, str(exc.detail)), + status_code=exc.status_code, + ) + + +_SAFE_TAGS = { + "a", "abbr", "article", "b", "blockquote", "br", "caption", "code", "dd", "del", "details", + "div", "dl", "dt", "em", "figcaption", "figure", "h1", "h2", "h3", "h4", "h5", "h6", + "hr", "i", "img", "li", "main", "mark", "ol", "p", "pre", "s", "section", "small", "span", + "strong", "sub", "summary", "sup", "table", "tbody", "td", "th", "thead", "tr", "u", "ul", +} +_VOID_TAGS = {"br", "hr", "img"} +_DROP_CONTENT_TAGS = {"script", "style", "iframe", "object", "embed", "svg", "math", "form", "template"} +_SAFE_ATTRS = {"alt", "title", "class", "id", "colspan", "rowspan", "width", "height", "scope", "dir", "lang"} + + +class _HTMLPreviewSanitizer(HTMLParser): + """Whitelist-only HTML serializer; scripts, active embeds and URL attrs are removed.""" + + def __init__(self): + super().__init__(convert_charrefs=True) + self.output: list[str] = [] + self.drop_depth = 0 + + def handle_starttag(self, tag, attrs): + tag = tag.lower() + if self.drop_depth: + if tag in _DROP_CONTENT_TAGS: + self.drop_depth += 1 + return + if tag in _DROP_CONTENT_TAGS: + self.drop_depth = 1 + return + if tag not in _SAFE_TAGS: + return + safe_attrs = [] + for name, value in attrs: + name = name.lower() + if value is None or name.startswith("on") or name not in _SAFE_ATTRS: + continue + safe_attrs.append(f' {name}="{html.escape(value, quote=True)}"') + self.output.append(f"<{tag}{''.join(safe_attrs)}>") + + def handle_startendtag(self, tag, attrs): + self.handle_starttag(tag, attrs) + if tag.lower() in _SAFE_TAGS and tag.lower() not in _VOID_TAGS: + self.handle_endtag(tag) + + def handle_endtag(self, tag): + tag = tag.lower() + if self.drop_depth: + if tag in _DROP_CONTENT_TAGS: + self.drop_depth -= 1 + return + if tag in _SAFE_TAGS and tag not in _VOID_TAGS: + self.output.append(f"") + + def handle_data(self, data): + if not self.drop_depth: + self.output.append(html.escape(data)) + + def handle_entityref(self, name): + if not self.drop_depth: + self.output.append(f"&{html.escape(name)};") + + def handle_charref(self, name): + if not self.drop_depth: + self.output.append(f"&#{html.escape(name)};") + + +def sanitize_html_preview(source: str) -> str: + parser = _HTMLPreviewSanitizer() + parser.feed(source) + parser.close() + return "" + "".join(parser.output) + "" diff --git a/apps/base/views.py b/apps/base/views.py index f52784d..b5ed711 100644 --- a/apps/base/views.py +++ b/apps/base/views.py @@ -1,4 +1,5 @@ import hashlib +import asyncio import hmac import os import uuid @@ -9,8 +10,7 @@ from typing import Optional, Tuple, Union from fastapi import APIRouter, Form, Request, UploadFile, File, Depends, HTTPException from pydantic import BaseModel, ValidationError -from starlette import status -from starlette.responses import Response +from starlette.responses import HTMLResponse, Response from tortoise.expressions import Case, F, Q, When from apps.base.auth import share_required_login @@ -39,6 +39,7 @@ from apps.base.utils import ( validate_expire_style, ) from apps.base.local_share import is_local_ref +from apps.base.mcp import sanitize_html_preview from core.response import APIResponse from core.settings import settings from core.storage import storages, FileStorageInterface @@ -194,6 +195,51 @@ async def post_file_metadata( return APIResponse(detail=build_file_metadata(file_code)) +@share_api.get("/preview/html/{code}", include_in_schema=False) +async def preview_html(code: str): + has, file_code = await get_code_file_by_code(code) + if not has or not isinstance(file_code, FileCodes): + raise HTTPException(status_code=404, detail="文件不存在") + if file_code.text is not None or (file_code.suffix or "").lower() not in {".html", ".htm"}: + raise HTTPException(status_code=415, detail="仅支持 HTML 文件预览") + preview_limit = 2 * 1024 * 1024 + if file_code.size > preview_limit: + raise HTTPException(status_code=413, detail="HTML 文件过大,无法在线预览") + download = await get_stored_download(file_code) + if download.path is not None: + if download.path.stat().st_size > preview_limit: + raise HTTPException(status_code=413, detail="HTML 文件过大,无法在线预览") + content = await asyncio.to_thread(download.path.read_bytes) + elif download.content is not None: + content = download.content + elif download.stream_factory is not None: + stream = download.stream_factory() + if hasattr(stream, "__aiter__"): + chunks = bytearray() + async for chunk in stream: + chunks.extend(chunk) + if len(chunks) > preview_limit: + raise HTTPException(status_code=413, detail="HTML 文件过大,无法在线预览") + content = bytes(chunks) + else: + raise HTTPException(status_code=500, detail="HTML 预览读取失败") + else: + raise HTTPException(status_code=500, detail="HTML 预览读取失败") + if len(content) > preview_limit: + raise HTTPException(status_code=413, detail="HTML 文件过大,无法在线预览") + safe_html = sanitize_html_preview(content.decode("utf-8-sig", errors="replace")) + return HTMLResponse( + safe_html, + headers={ + "Content-Security-Policy": "sandbox; default-src 'none'; img-src data:; style-src 'unsafe-inline'; font-src data:; base-uri 'none'; form-action 'none'", + "X-Content-Type-Options": "nosniff", + "Cache-Control": "private, no-store", + "Content-Disposition": 'inline; filename="preview.html"', + }, + background=download.background, + ) + + @share_api.get("/select/") async def get_code_file(code: str, ip: str = Depends(ip_limit["error"])): file_storage: FileStorageInterface = storages[settings.file_storage]() diff --git a/core/settings.py b/core/settings.py index f66905e..26a1f08 100644 --- a/core/settings.py +++ b/core/settings.py @@ -48,6 +48,7 @@ DEFAULT_CONFIG: dict[str, Any] = { "jwt_secret": "", "admin_session_expire": ADMIN_SESSION_EXPIRE_DEFAULT, "open_upload": 1, + "mcp_require_token": 0, "upload_size": 1024 * 1024 * 10, "allowed_file_types": ["*"], "expire_style": ["day", "hour", "minute", "forever", "count"], diff --git a/main.py b/main.py index 98e5c06..c4806da 100644 --- a/main.py +++ b/main.py @@ -28,6 +28,7 @@ from apps.base.tasks import ( delete_expire_files, ) from apps.base.views import share_api, chunk_api, presign_api +from apps.base.mcp import router as mcp_router from core.database import db_startup_lock, get_db_config, init_db from core.errors import StorageError from core.logger import get_log_level_name, is_access_log_enabled, logger @@ -139,6 +140,7 @@ app.include_router(presign_api) app.include_router(presign_api, prefix="/api") app.include_router(admin_api) app.include_router(pages_router) +app.include_router(mcp_router) # 404 时返回主题首页(index 兼任 exception handler 与 GET / 路由) app.add_exception_handler(404, not_found_handler) diff --git a/tests/test_mcp_upload.py b/tests/test_mcp_upload.py new file mode 100644 index 0000000..93eb279 --- /dev/null +++ b/tests/test_mcp_upload.py @@ -0,0 +1,117 @@ +import base64 + +import pytest + +from apps.base.models import KeyValue +from apps.base.mcp import sanitize_html_preview + + +def test_html_preview_sanitizer_removes_active_content_and_event_handlers(): + output = sanitize_html_preview( + '

Hello

' + 'link' + ) + assert "Hello" in output + assert "