| 1 | import os |
| 2 | import sys |
| 3 | |
| 4 | models_dir = os.path.dirname(os.path.abspath(__file__)) |
| 5 | backend_dir = os.path.dirname(models_dir) |
| 6 | if backend_dir not in sys.path: |
| 7 | sys.path.insert(0, backend_dir) |
| 8 | |
| 9 | import logging |
| 10 | |
| 11 | try: |
| 12 | from models.llm_gpt import GPT |
| 13 | from models.llm_gemini import Gemini |
| 14 | from models.llm_deepseek import DeepSeek |
| 15 | from models.llm_dashscope import QwenLLM |
| 16 | from models.vlm_dashscope import QwenVLClient |
| 17 | except ImportError: |
| 18 | from llm_gpt import GPT |
| 19 | from llm_gemini import Gemini |
| 20 | from llm_deepseek import DeepSeek |
| 21 | from llm_dashscope import QwenLLM |
| 22 | from vlm_dashscope import QwenVLClient |
| 23 | |
| 24 | from config import Config |
| 25 | |
| 26 | logger = logging.getLogger(__name__) |
| 27 | |
| 28 | class LLM: |
| 29 | def __init__(self, gemini_base_url="", gemini_api_key="", gpt_base_url="", gpt_api_key="", deepseek_base_url="", deepseek_api_key="", dashscope_api_key=""): |
| 30 | self._gemini_base_url = gemini_base_url or Config.GOOGLE_GEMINI_BASE_URL |
| 31 | self._gemini_api_key = gemini_api_key or Config.GEMINI_API_KEY |
| 32 | self._gpt_base_url = gpt_base_url or Config.OPENAI_BASE_URL |
| 33 | self._gpt_api_key = gpt_api_key or Config.OPENAI_API_KEY |
| 34 | self._deepseek_base_url = deepseek_base_url or Config.DEEPSEEK_BASE_URL |
| 35 | self._deepseek_api_key = deepseek_api_key or Config.DEEPSEEK_API_KEY |
| 36 | self._dashscope_api_key = dashscope_api_key or Config.DASHSCOPE_API_KEY |
| 37 | |
| 38 | self._gemini_client = None |
| 39 | self._gpt_client = None |
| 40 | self._deepseek_client = None |
| 41 | self._dashscope_client = None |
| 42 | self._dashscope_vl_client = None |
| 43 | |
| 44 | @property |
| 45 | def gemini_client(self): |
| 46 | if self._gemini_client is None: |
| 47 | self._gemini_client = Gemini( |
| 48 | base_url=self._gemini_base_url, |
| 49 | api_key=self._gemini_api_key, |
| 50 | ) |
| 51 | return self._gemini_client |
| 52 | |
| 53 | @property |
| 54 | def gpt_client(self): |
| 55 | if self._gpt_client is None: |
| 56 | self._gpt_client = GPT( |
| 57 | base_url=self._gpt_base_url, |
| 58 | api_key=self._gpt_api_key, |
| 59 | proxy=Config.provider_proxy("openai"), |
| 60 | ) |
| 61 | return self._gpt_client |
| 62 | |
| 63 | @property |
| 64 | def deepseek_client(self): |
| 65 | if self._deepseek_client is None: |
| 66 | self._deepseek_client = DeepSeek( |
| 67 | base_url=self._deepseek_base_url, |
| 68 | api_key=self._deepseek_api_key, |
| 69 | ) |
| 70 | return self._deepseek_client |
| 71 | |
| 72 | @property |
| 73 | def dashscope_client(self): |
| 74 | if self._dashscope_client is None: |
| 75 | self._dashscope_client = QwenLLM(api_key=self._dashscope_api_key) |
| 76 | return self._dashscope_client |
| 77 | |
| 78 | @property |
| 79 | def dashscope_vl_client(self): |
| 80 | if self._dashscope_vl_client is None: |
| 81 | self._dashscope_vl_client = QwenVLClient(api_key=self._dashscope_api_key) |
| 82 | return self._dashscope_vl_client |
| 83 | |
| 84 | def full_to_half(self, text): |
| 85 | if not isinstance(text, str): |
| 86 | return text |
| 87 | |
| 88 | translation_table = {0x3000: 0x0020} |
| 89 | for i in range(65281, 65375): |
| 90 | translation_table[i] = i - 65248 |
| 91 | |
| 92 | return text.translate(translation_table) |
| 93 | |
| 94 | def query(self, prompt, image_urls=[], model="qwen3.6-max-preview", safe_content=True, task_id=None, web_search=False): |
| 95 | """ |
| 96 | Query the LLM with a prompt and optional image URLs. |
| 97 | Selects the backend (GPT or Gemini) based on the model name. |
| 98 | |
| 99 | :param web_search: Enable web search for supported providers |
| 100 | """ |
| 101 | if safe_content: |
| 102 | prompt = self.full_to_half(prompt) |
| 103 | |
| 104 | if not model: |
| 105 | model = "qwen3.6-max-preview" |
| 106 | |
| 107 | if Config.PRINT_MODEL_INPUT: |
| 108 | lines = [ |
| 109 | "---- LLM QUERY REQUEST ----", |
| 110 | f"Model: {model}", |
| 111 | ] |
| 112 | if task_id: |
| 113 | lines.append(f"Task ID: {task_id}") |
| 114 | if image_urls: |
| 115 | lines.append(f"Images: {len(image_urls)}") |
| 116 | lines.extend(f" - {u}" for u in image_urls) |
| 117 | lines.extend([ |
| 118 | f"Web Search: {web_search}", |
| 119 | f"Prompt: {prompt[:200]}{'...' if len(prompt) > 200 else ''}", |
| 120 | "-" * 30, |
| 121 | ]) |
| 122 | logger.info("\n%s", "\n".join(lines)) |
| 123 | |
| 124 | result = "" |
| 125 | model_lower = model.lower() |
| 126 | if model_lower.startswith("gemini"): |
| 127 | result = self.gemini_client.query(prompt, image_urls=image_urls, model=model) |
| 128 | elif "gpt" in model_lower: |
| 129 | # OpenAI series models |
| 130 | result = self.gpt_client.query(prompt, image_urls=image_urls, model=model, web_search=web_search) |
| 131 | elif "kimi" in model_lower or "qwen3.6-plus" in model_lower or "qwen3.6-flash" in model_lower or "vl" in model_lower: |
| 132 | # DashScope VLM models (using MultiModalConversation API) |
| 133 | result = self.dashscope_vl_client.chat(text=prompt, images=image_urls, model=model, stream=False) |
| 134 | elif "deepseek-v3.2" in model_lower: |
| 135 | # DeepSeek v3.2 (通过 DashScope Generation API) |
| 136 | result = self.dashscope_client.query(prompt, image_urls=image_urls, model=model, web_search=web_search) |
| 137 | elif model_lower.startswith("deepseek") and "v3.2" not in model_lower: |
| 138 | # Original DeepSeek provider |
| 139 | result = self.deepseek_client.query(prompt, image_urls=image_urls, model=model, web_search=web_search) |
| 140 | else: |
| 141 | # Default to Qwen models / deepseek-v3.2 via DashScope Generation API |
| 142 | result = self.dashscope_client.query(prompt, image_urls=image_urls, model=model, web_search=web_search) |
| 143 | |
| 144 | if safe_content: |
| 145 | result = self.full_to_half(result) |
| 146 | |
| 147 | # Remove empty lines |
| 148 | return '\n'.join([line for line in result.split('\n') if line.strip() != '']) |
| 149 |