| 1 | import os |
| 2 | from typing import List, Optional |
| 3 | from .config import Config |
| 4 | |
| 5 | try: |
| 6 | from .vlm_dashscope import QwenVLClient |
| 7 | except ImportError: |
| 8 | from vlm_dashscope import QwenVLClient |
| 9 | |
| 10 | class VLM: |
| 11 | def __init__(self, |
| 12 | dashscope_api_key: Optional[str] = None, |
| 13 | dashscope_base_url: Optional[str] = None): |
| 14 | """ |
| 15 | Unified VLM (Vision Language Model) Client |
| 16 | Routes asset-analysis VLM requests to DashScope (Qwen/Qwen-Omni). |
| 17 | """ |
| 18 | dashscope_key = dashscope_api_key or Config.DASHSCOPE_API_KEY |
| 19 | |
| 20 | self.dashscope_client = ( |
| 21 | QwenVLClient( |
| 22 | api_key=dashscope_key, |
| 23 | base_url=dashscope_base_url or Config.DASHSCOPE_BASE_URL |
| 24 | ) |
| 25 | if dashscope_key else None |
| 26 | ) |
| 27 | |
| 28 | def query(self, |
| 29 | prompt: str, |
| 30 | image_paths: Optional[List[str]] = None, |
| 31 | model: Optional[str] = None, |
| 32 | session_id: Optional[str] = None, |
| 33 | video_paths: Optional[List[str]] = None) -> str: |
| 34 | selected_model = (model or "").strip() |
| 35 | if not selected_model: |
| 36 | raise RuntimeError("DashScope VLM model must be explicitly selected.") |
| 37 | |
| 38 | if Config.PRINT_MODEL_INPUT: |
| 39 | print("---- VLM REQUEST ----") |
| 40 | print(f"Prompt: {prompt}") |
| 41 | if image_paths: |
| 42 | print(f"Images: {len(image_paths)}") |
| 43 | for p in image_paths: |
| 44 | if p.startswith("data:"): |
| 45 | print(f" - [Base64图片]") |
| 46 | else: |
| 47 | print(f" - {p}") |
| 48 | if video_paths: |
| 49 | print(f"Videos: {len(video_paths)}") |
| 50 | for p in video_paths: |
| 51 | print(f" - {p}") |
| 52 | print(f"Model: {selected_model}") |
| 53 | if session_id: |
| 54 | print(f"Session ID: {session_id}") |
| 55 | print("-" * 30) |
| 56 | |
| 57 | if self.dashscope_client is None: |
| 58 | raise RuntimeError("DashScope VLM API key is not configured.") |
| 59 | |
| 60 | image_urls = [self._to_dashscope_file_url(path, allow_data_url=True) for path in image_paths or []] |
| 61 | video_urls = [self._to_dashscope_file_url(path, allow_data_url=False) for path in video_paths or []] |
| 62 | return self.dashscope_client.chat( |
| 63 | text=prompt, |
| 64 | images=image_urls, |
| 65 | videos=video_urls, |
| 66 | model=selected_model, |
| 67 | stream=False, |
| 68 | ) |
| 69 | |
| 70 | def _to_dashscope_file_url(self, path: str, allow_data_url: bool) -> str: |
| 71 | if path.startswith("data:"): |
| 72 | if not allow_data_url: |
| 73 | raise ValueError("DashScope video input does not support data URLs in this adapter.") |
| 74 | |
| 75 | import base64 as b64 |
| 76 | import tempfile |
| 77 | |
| 78 | try: |
| 79 | header, b64_data = path.split(",", 1) |
| 80 | mime_type = header.split(";")[0].replace("data:", "") |
| 81 | image_data = b64.b64decode(b64_data) |
| 82 | suffix = f".{mime_type.split('/')[-1]}" if "/" in mime_type else ".png" |
| 83 | with tempfile.NamedTemporaryFile(suffix=suffix, delete=False) as tmp: |
| 84 | tmp.write(image_data) |
| 85 | temp_path = tmp.name |
| 86 | return f"file://{os.path.abspath(temp_path)}" |
| 87 | except Exception as e: |
| 88 | print(f"Error processing base64 image: {e}") |
| 89 | raise ValueError(f"无法解析 base64 图片: {e}") |
| 90 | |
| 91 | if path.startswith("http") or path.startswith("file://"): |
| 92 | return path |
| 93 | |
| 94 | return f"file://{os.path.abspath(path)}" |
| 95 |