From cb80d5cee73502a4f7eaed250a7eae1fc2251b82 Mon Sep 17 00:00:00 2001 From: MT-Fire <798521692@qq.com> Date: Wed, 4 Mar 2026 18:00:14 +0800 Subject: [PATCH] =?UTF-8?q?feat:=20=E6=96=B0=E5=A2=9E=20oss=20=E4=B8=8E?= =?UTF-8?q?=E6=99=BA=E6=99=AE=E5=9B=BE=E7=89=87=E8=A7=A3=E6=9E=90=E5=8A=9F?= =?UTF-8?q?=E8=83=BD?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- server/api/chat_routes.py | 19 ++- server/requirements.txt | 2 + server/utils/glm_adapter.py | 85 ++++++++++++- server/utils/oss_uploader.py | 227 +++++++++++++++++++++++++++++++++++ server/utils/test_oss_glm.py | 89 ++++++++++++++ 5 files changed, 415 insertions(+), 7 deletions(-) create mode 100644 server/utils/oss_uploader.py create mode 100644 server/utils/test_oss_glm.py diff --git a/server/api/chat_routes.py b/server/api/chat_routes.py index ca91c7f..25dbc90 100644 --- a/server/api/chat_routes.py +++ b/server/api/chat_routes.py @@ -35,8 +35,12 @@ from pathlib import Path sys.path.append(str(Path(__file__).parent.parent)) from models.chat_models import ChatRequest, ModelInfo -from utils.helpers import (extract_delta_content, format_api_response, - generate_unique_id, get_current_timestamp) +from utils.helpers import ( + extract_delta_content, + format_api_response, + generate_unique_id, + get_current_timestamp, +) from utils.logger import log_error, log_exception, log_info # 模拟数据库 - 实际应用中应使用持久化存储 @@ -1030,13 +1034,18 @@ async def upload_file_handler(file: UploadFile = File(...)): unique_filename = f"{int(datetime.utcnow().timestamp())}_{generate_unique_id()}{file_extension}" file_path = upload_dir / unique_filename - # 保存文件 + # 保存文件到本地（临时缓存） + content = await file.read() with open(file_path, "wb") as f: - content = await file.read() f.write(content) + # 文件关闭后再上传到 OSS + from utils.oss_uploader import upload_file as oss_upload + + oss_result = oss_upload(str(file_path)) + file_url = oss_result["url"] + # 返回文件信息 - file_url = f"http://localhost:8000/uploads/{unique_filename}" result = { "url": file_url, "name": file.filename, diff --git a/server/requirements.txt b/server/requirements.txt index 2fa7527..f580542 100644 --- a/server/requirements.txt +++ b/server/requirements.txt @@ -9,3 +9,5 @@ typing-extensions==4.12.2 # 路线二：阿里云文档智能解析（doc/docx/pdf） llama-index-core>=0.10.0 llama-index-readers-dashscope>=0.1.0 +# 阿里云 OSS 上传 +alibabacloud-oss-v2 diff --git a/server/utils/glm_adapter.py b/server/utils/glm_adapter.py index 3040cc3..157c8a9 100644 --- a/server/utils/glm_adapter.py +++ b/server/utils/glm_adapter.py @@ -47,7 +47,7 @@ def get_client(): # ── 模型映射 ────────────────────────────────────────────────────────── DEFAULT_TEXT_MODEL = "glm-4.5-Air" # glm-4.6 文本统一模型 -DEFAULT_VISION_MODEL = "glm-4.5-Air" +DEFAULT_VISION_MODEL = "glm-4.6v" MODEL_MAP = { "qwen-max": "glm-4.5-Air", @@ -61,7 +61,12 @@ MODEL_MAP = { def resolve_model(model: str, has_vision: bool = False) -> str: if model.startswith("glm"): return model - return MODEL_MAP.get(model, DEFAULT_TEXT_MODEL) + mapped = MODEL_MAP.get(model, DEFAULT_TEXT_MODEL) + # 当消息包含图片时，强制使用视觉模型 + if has_vision and mapped != DEFAULT_VISION_MODEL: + print(f"[GLM] 检测到图片，模型从 {mapped} 切换为 {DEFAULT_VISION_MODEL}") + return DEFAULT_VISION_MODEL + return mapped # ── 文件上传（含 file_id 缓存）─────────────────────────────────────── @@ -249,6 +254,44 @@ async def glm_stream_generator( f"[GLM] 流式请求：model={actual_model} vision={has_vision} " f"web_search={web_search} thinking={deep_thinking}" ) + # ── 调试：打印发送给 GLM 的完整消息结构 ── + for i, msg in enumerate(glm_msgs): + role = msg.get("role", "?") + content = msg.get("content", "") + if isinstance(content, list): + for j, part in enumerate(content): + if not isinstance(part, dict): + print(f"[GLM-DEBUG] msg[{i}].content[{j}]: {type(part).__name__}") + continue + part_type = part.get("type", "?") + if part_type == "image_url": + img_val = part.get("image_url", "") + img_url = ( + img_val.get("url", "") + if isinstance(img_val, dict) + else str(img_val) + ) + display = img_url[:120] + "..." if len(img_url) > 120 else img_url + print( + f"[GLM-DEBUG] msg[{i}].content[{j}]: type=image_url, url={display}" + ) + elif part_type == "text": + preview = (part.get("text", "") or "")[:100] + print( + f"[GLM-DEBUG] msg[{i}].content[{j}]: type=text, text={preview}" + ) + else: + print(f"[GLM-DEBUG] msg[{i}].content[{j}]: {part}") + else: + print(f"[GLM-DEBUG] msg[{i}]: role={role}, content={str(content)[:150]}") + if extra_kwargs: + print(f"[GLM-DEBUG] extra_kwargs={extra_kwargs}") + # 原始 JSON 转储（用于排查结构问题） + import json as _json + + print( + f"[GLM-RAW] messages={_json.dumps(glm_msgs, ensure_ascii=False, default=str)[:2000]}" + ) chunk_queue: queue.Queue = queue.Queue(maxsize=128) @@ -367,6 +410,44 @@ def glm_chat_sync( client = get_client() print(f"[GLM] 非流式请求：model={actual_model}") + # ── 调试：打印发送给 GLM 的完整消息结构 ── + for i, msg in enumerate(glm_msgs): + role = msg.get("role", "?") + content = msg.get("content", "") + if isinstance(content, list): + for j, part in enumerate(content): + if not isinstance(part, dict): + print(f"[GLM-DEBUG] msg[{i}].content[{j}]: {type(part).__name__}") + continue + part_type = part.get("type", "?") + if part_type == "image_url": + img_val = part.get("image_url", "") + img_url = ( + img_val.get("url", "") + if isinstance(img_val, dict) + else str(img_val) + ) + display = img_url[:120] + "..." if len(img_url) > 120 else img_url + print( + f"[GLM-DEBUG] msg[{i}].content[{j}]: type=image_url, url={display}" + ) + elif part_type == "text": + preview = (part.get("text", "") or "")[:100] + print( + f"[GLM-DEBUG] msg[{i}].content[{j}]: type=text, text={preview}" + ) + else: + print(f"[GLM-DEBUG] msg[{i}].content[{j}]: {part}") + else: + print(f"[GLM-DEBUG] msg[{i}]: role={role}, content={str(content)[:150]}") + if extra_kwargs: + print(f"[GLM-DEBUG] extra_kwargs={extra_kwargs}") + # 原始 JSON 转储（用于排查结构问题） + import json as _json + + print( + f"[GLM-RAW] messages={_json.dumps(glm_msgs, ensure_ascii=False, default=str)[:2000]}" + ) resp = client.chat.completions.create( model=actual_model, messages=glm_msgs, diff --git a/server/utils/oss_uploader.py b/server/utils/oss_uploader.py new file mode 100644 index 0000000..e47706f --- /dev/null +++ b/server/utils/oss_uploader.py @@ -0,0 +1,227 @@ +""" +阿里云 OSS 简单上传工具 +参考文档: https://help.aliyun.com/zh/oss/user-guide/simple-upload + +支持: + - 上传本地文件 + - 上传字节数据 / 字符串 + - 上传文件流（file-like object） + - 自动根据文件后缀生成 OSS 对象路径 + - 返回可公开访问的 URL（需确保 Bucket 已开启公共读或已配置签名） + +前置条件: + 1. pip install alibabacloud-oss-v2 + 2. 在 .env 中配置以下变量: + OSS_ACCESS_KEY_ID=<你的 AccessKey ID> + OSS_ACCESS_KEY_SECRET=<你的 AccessKey Secret> + OSS_BUCKET_NAME=<存储空间名称> + OSS_ENDPOINT= + OSS_REGION=<地域，如 cn-hangzhou> + OSS_URL_PREFIX=<可选，自定义域名前缀，如 https://cdn.example.com> +""" + +import os +import uuid +import mimetypes +from datetime import datetime +from pathlib import Path +from typing import Optional, Union, BinaryIO + +import alibabacloud_oss_v2 as oss +from dotenv import load_dotenv + +# ── 加载环境变量 ────────────────────────────────────────────── +load_dotenv() + +# AccessKey 从系统环境变量读取（~/.bashrc 中 export 设置） +OSS_ACCESS_KEY_ID = os.environ.get("OSS_ACCESS_KEY_ID", "") +OSS_ACCESS_KEY_SECRET = os.environ.get("OSS_ACCESS_KEY_SECRET", "") +# 以下配置从 .env 文件读取 +OSS_BUCKET_NAME = os.getenv("OSS_BUCKET_NAME", "") +OSS_ENDPOINT = os.getenv("OSS_ENDPOINT", "") +OSS_REGION = os.getenv("OSS_REGION", "") +# 可选：自定义域名前缀，用于拼接返回的公开 URL +OSS_URL_PREFIX = os.getenv("OSS_URL_PREFIX", "") + + +def _get_client() -> oss.Client: + """创建并返回 OSS 客户端实例""" + credentials_provider = oss.credentials.StaticCredentialsProvider( + access_key_id=OSS_ACCESS_KEY_ID, + access_key_secret=OSS_ACCESS_KEY_SECRET, + ) + cfg = oss.config.load_default() + cfg.credentials_provider = credentials_provider + cfg.region = OSS_REGION + cfg.endpoint = OSS_ENDPOINT + return oss.Client(cfg) + + +def _generate_object_key(filename: str, prefix: str = "uploads") -> str: + """ + 根据文件名生成唯一的 OSS 对象 Key + 格式: {prefix}/{日期}/{uuid}_{原始文件名} + """ + date_str = datetime.now().strftime("%Y%m%d") + unique_id = uuid.uuid4().hex[:8] + safe_name = Path(filename).name # 只取文件名，去掉路径 + return f"{prefix}/{date_str}/{unique_id}_{safe_name}" + + +def _build_url(object_key: str) -> str: + """根据对象 Key 构建可访问的 URL""" + if OSS_URL_PREFIX: + return f"{OSS_URL_PREFIX.rstrip('/')}/{object_key}" + # 默认使用 Bucket 域名拼接 + endpoint = OSS_ENDPOINT.replace("https://", "").replace("http://", "") + return f"https://{OSS_BUCKET_NAME}.{endpoint}/{object_key}" + + +def upload_file( + file_path: str, + object_key: Optional[str] = None, + prefix: str = "uploads", +) -> dict: + """ + 上传本地文件到 OSS + + 参数: + file_path: 本地文件的绝对路径 + object_key: 自定义 OSS 对象名称，为 None 则自动生成 + prefix: 对象 Key 的前缀目录 + + 返回: + 包含上传结果的字典: + { + "url": "文件访问地址", + "object_key": "OSS 对象路径", + "etag": "ETag", + "status_code": 200, + } + """ + if not os.path.isfile(file_path): + raise FileNotFoundError(f"文件不存在: {file_path}") + + filename = os.path.basename(file_path) + if object_key is None: + object_key = _generate_object_key(filename, prefix) + + # 检测文件类型，设置 Content-Type + content_type, _ = mimetypes.guess_type(file_path) + + client = _get_client() + result = client.put_object_from_file( + oss.PutObjectRequest( + bucket=OSS_BUCKET_NAME, + key=object_key, + content_type=content_type, + ), + file_path, + ) + + return { + "url": _build_url(object_key), + "object_key": object_key, + "etag": result.etag, + "status_code": result.status_code, + } + + +def upload_bytes( + data: Union[bytes, str], + filename: str, + object_key: Optional[str] = None, + prefix: str = "uploads", + content_type: Optional[str] = None, +) -> dict: + """ + 上传字节数据或字符串到 OSS + + 参数: + data: 要上传的数据（bytes 或 str） + filename: 用于生成 Key 的文件名（如 "report.txt"） + object_key: 自定义 OSS 对象名称，为 None 则自动生成 + prefix: 对象 Key 的前缀目录 + content_type: 自定义 Content-Type + + 返回: + 包含上传结果的字典 + """ + if isinstance(data, str): + data = data.encode("utf-8") + + if object_key is None: + object_key = _generate_object_key(filename, prefix) + + if content_type is None: + content_type, _ = mimetypes.guess_type(filename) + + client = _get_client() + result = client.put_object( + oss.PutObjectRequest( + bucket=OSS_BUCKET_NAME, + key=object_key, + body=data, + content_type=content_type, + ) + ) + + return { + "url": _build_url(object_key), + "object_key": object_key, + "etag": result.etag, + "status_code": result.status_code, + } + + +def upload_fileobj( + fileobj: BinaryIO, + filename: str, + object_key: Optional[str] = None, + prefix: str = "uploads", + content_type: Optional[str] = None, +) -> dict: + """ + 上传文件流（file-like object）到 OSS + + 参数: + fileobj: 文件流对象（如 open(path, 'rb') 或 FastAPI 的 UploadFile.file） + filename: 用于生成 Key 的文件名 + object_key: 自定义 OSS 对象名称，为 None 则自动生成 + prefix: 对象 Key 的前缀目录 + content_type: 自定义 Content-Type + + 返回: + 包含上传结果的字典 + """ + data = fileobj.read() + return upload_bytes( + data=data, + filename=filename, + object_key=object_key, + prefix=prefix, + content_type=content_type, + ) + + +# ──────────────────────────────────────────────────────────────── +# 命令行入口：python -m utils.oss_uploader --file <路径> +# ──────────────────────────────────────────────────────────────── +if __name__ == "__main__": + import argparse + + parser = argparse.ArgumentParser(description="阿里云 OSS 简单上传工具") + parser.add_argument("--file", required=True, help="要上传的本地文件路径") + parser.add_argument("--key", default=None, help="自定义 OSS 对象路径（可选）") + parser.add_argument( + "--prefix", default="uploads", help="对象 Key 前缀（默认: uploads）" + ) + args = parser.parse_args() + + result = upload_file(args.file, object_key=args.key, prefix=args.prefix) + + print("✅ 上传成功!") + print(f" 访问地址: {result['url']}") + print(f" 对象路径: {result['object_key']}") + print(f" ETag: {result['etag']}") + print(f" 状态码: {result['status_code']}") diff --git a/server/utils/test_oss_glm.py b/server/utils/test_oss_glm.py new file mode 100644 index 0000000..cd0e56a --- /dev/null +++ b/server/utils/test_oss_glm.py @@ -0,0 +1,89 @@ +""" +测试脚本：上传文件到 OSS → 获取 URL → 发送给 GLM 进行识别 + +用法: + cd server + source ~/.bashrc && source .venv/bin/activate + python -m utils.test_oss_glm --file <本地文件路径> [--prompt "描述一下这张图片"] +""" + +import argparse +import sys +from pathlib import Path + +# 确保 server 目录在 sys.path +sys.path.insert(0, str(Path(__file__).resolve().parent.parent)) + +from utils.oss_uploader import upload_file +from utils.glm_adapter import glm_chat_sync + + +def main(): + parser = argparse.ArgumentParser(description="上传文件到 OSS 并让 GLM 读取") + parser.add_argument("--file", required=True, help="要上传的本地文件路径") + parser.add_argument( + "--prompt", default="请描述一下这张图片的内容", help="发给 GLM 的提示词" + ) + parser.add_argument( + "--model", default="glm-4.6v", help="GLM 模型名称（默认: glm-4.6v）" + ) + args = parser.parse_args() + + # ── 第一步：上传文件到 OSS ──────────────────────────────── + file_path = args.file + if not Path(file_path).exists(): + print(f"❌ 文件不存在: {file_path}") + sys.exit(1) + + print(f"📤 正在上传文件: {file_path}") + oss_result = upload_file(file_path) + file_url = oss_result["url"] + print(f"✅ 上传成功!") + print(f" URL: {file_url}") + print() + + # ── 第二步：构建消息，把 URL 发给 GLM ────────────────────── + messages = [ + { + "role": "user", + "content": [ + { + "type": "image_url", + "image_url": {"url": file_url}, + }, + { + "type": "text", + "text": args.prompt, + }, + ], + } + ] + + print(f"🤖 正在请求 GLM ({args.model}) 识别图片...") + print(f" 提示词: {args.prompt}") + print() + + result = glm_chat_sync( + messages=messages, + model=args.model, + temperature=0.7, + max_tokens=2048, + ) + + print("─" * 60) + print("📝 GLM 回复:") + print("─" * 60) + print(result["content"]) + print("─" * 60) + + if result.get("usage"): + usage = result["usage"] + print( + f"\n📊 Token 用量: 输入 {usage['promptTokens']} | " + f"输出 {usage['completionTokens']} | " + f"总计 {usage['totalTokens']}" + ) + + +if __name__ == "__main__": + main()