From b4c6faab868bedf72ae9b40554d129984c67d57b Mon Sep 17 00:00:00 2001 From: martsforever Date: Sat, 28 Mar 2026 23:33:40 +0800 Subject: [PATCH] =?UTF-8?q?feat:=20=E7=BB=9F=E4=B8=80=E7=AE=A1=E7=90=86?= =?UTF-8?q?=E6=A8=A1=E5=9E=8B=E6=9C=8D=E5=8A=A1=E9=85=8D=E7=BD=AE=E4=BF=A1?= =?UTF-8?q?=E6=81=AF?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .env.example | 5 +++ app/config/ai_configs.py | 84 ++++++++++++++++++++++++++++++++++++++++ app/config/env.py | 5 +++ app/utils/llm_utils.py | 29 ++++++++++++++ main.py | 22 ++++++----- 5 files changed, 135 insertions(+), 10 deletions(-) create mode 100644 app/config/ai_configs.py create mode 100644 app/utils/llm_utils.py diff --git a/.env.example b/.env.example index 454468a..c53a1a5 100644 --- a/.env.example +++ b/.env.example @@ -1 +1,6 @@ +LLM_KEY_LOCAL=123 # 本地模型key +LLM_KEY_HUOSHAN=dc7979dc-af60-4700-9f71-91f576b1c947 # 火山引擎模型服务平台key +LLM_KEY_BAILIAN=sk-248f811295914adcad837add0a5971d0 # 阿里云百炼模型服务平台key +LLM_KEY_DEEPSEEK=sk-a89d0ff9421a43fca5f0xxxxxxxxxxxx # Deepseek模型服务平台key + SERVER_PORT = 7004 # 服务启动端口 diff --git a/app/config/ai_configs.py b/app/config/ai_configs.py new file mode 100644 index 0000000..6bba938 --- /dev/null +++ b/app/config/ai_configs.py @@ -0,0 +1,84 @@ +from app.config.env import env + +ai_configs = { + # /*---------------------------------------local-------------------------------------------*/ + "local": { + "model": "deepseek-r1-distill-qwen-7b", + "url": "http://127.0.0.1:1234/v1/chat/completions", + "key": env.llm_key_local + }, + "local_glm": { + "model": "glm-4-9b-0414", + "url": "http://127.0.0.1:1234/v1/chat/completions", + "key": env.llm_key_local + }, + "local_embedding": { + "model": "text-embedding-text2vec-large-chinese", + "url": "http://127.0.0.1:1234/v1/embeddings", + "key": env.llm_key_local + }, + # /*---------------------------------------deepseek-------------------------------------------*/ + 'deepseek-v3': { + 'model': 'deepseek-chat', + 'url': 'https://api.deepseek.com/chat/completions', + 'key': env.llm_key_deepseek, + }, + 'deepseek-r1': { + 'model': 'deepseek-reasoner', + 'url': 'https://api.deepseek.com/chat/completions', + 'key': env.llm_key_deepseek, + }, + # /*---------------------------------------huoshan-------------------------------------------*/ + 'huoshan-deepseek-v3': { + 'model': 'deepseek-v3-250324', + 'url': 'https://ark.cn-beijing.volces.com/api/v3/chat/completions', + 'key': env.llm_key_huoshan, + }, + 'huoshan-deepseek-r1': { + 'model': 'deepseek-r1-distill-qwen-7b-250120', + 'url': 'https://ark.cn-beijing.volces.com/api/v3/chat/completions', + 'key': env.llm_key_huoshan, + }, + 'huoshan-doubao': { + 'model': 'doubao-1-5-lite-32k-250115', + 'url': 'https://ark.cn-beijing.volces.com/api/v3/chat/completions', + 'key': env.llm_key_huoshan, + }, + 'huoshan-think-pro': { + 'model': 'doubao-1-5-thinking-pro-250415', + 'url': 'https://ark.cn-beijing.volces.com/api/v3/chat/completions', + 'key': env.llm_key_huoshan, + }, + 'huoshan-doubao-seed': { + 'model': 'doubao-seed-1-6-250615', + 'url': 'https://ark.cn-beijing.volces.com/api/v3/chat/completions', + 'key': env.llm_key_huoshan, + }, + 'doubao-vision-lite': { + 'model': 'doubao-1.5-vision-lite-250315', + 'url': 'https://ark.cn-beijing.volces.com/api/v3/chat/completions', + 'key': env.llm_key_huoshan, + }, + + 'huoshan-embedding-240715': { + "model": 'doubao-embedding-text-240715', + "url": 'https://ark.cn-beijing.volces.com/api/v3/embeddings', + 'key': env.llm_key_huoshan, + }, + # /*---------------------------------------bailian-------------------------------------------*/ + 'bailian-qwen-turbo': { + 'model': 'qwen-turbo', + 'url': 'https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions', + 'key': env.llm_key_bailian, + }, + 'bailian-qwen3-max': { + 'model': 'qwen3-max', + 'url': 'https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions', + 'key': env.llm_key_bailian, + }, + 'bailian-embedding': { + 'model': 'text-embedding-v4', + 'url': 'https://dashscope.aliyuncs.com/compatible-mode/v1/embeddings', + 'key': env.llm_key_bailian, + }, +} diff --git a/app/config/env.py b/app/config/env.py index ce95852..b4b0a75 100644 --- a/app/config/env.py +++ b/app/config/env.py @@ -3,6 +3,11 @@ from pydantic import Field class EnvSettings(BaseSettings): + llm_key_local: str = Field(..., env="LLM_KEY_LOCAL") + llm_key_huoshan: str = Field(..., env="LLM_KEY_HUOSHAN") + llm_key_bailian: str = Field(..., env="LLM_KEY_BAILIAN") + llm_key_deepseek: str = Field(..., env="LLM_KEY_DEEPSEEK") + server_port: int = Field(..., env="SERVER_PORT") class Config: diff --git a/app/utils/llm_utils.py b/app/utils/llm_utils.py new file mode 100644 index 0000000..c2ea6df --- /dev/null +++ b/app/utils/llm_utils.py @@ -0,0 +1,29 @@ +from langchain.chat_models import init_chat_model + +from app.config.ai_configs import ai_configs + + +def create_llm( + platform_code='huoshan-doubao', + temperature=0.5, + disable_streaming=False, +): + _ai_config = ai_configs.get(platform_code) + + if _ai_config is None: + raise Exception('Unknown platform code', platform_code) + + return init_chat_model( + # 必须填 openai（兼容协议） + model_provider="openai", + # 你的模型名 + model=_ai_config.get('model'), + # 你的自定义 API URL + base_url=_ai_config.get('url').replace("chat/completions", ""), + # 本地模型随便填 + api_key=_ai_config.get('key'), + # 随机性 + temperature=temperature, + # 禁用流式响应 + disable_streaming=disable_streaming, + ) diff --git a/main.py b/main.py index 926d83a..4c2f419 100644 --- a/main.py +++ b/main.py @@ -17,6 +17,7 @@ from starlette.staticfiles import StaticFiles from app.config.env import env from app.utils.get_local_ips import get_local_ips +from app.utils.llm_utils import create_llm @asynccontextmanager @@ -91,17 +92,18 @@ async def redirect_root_to_docs(): @app.get("/chat") async def chat(): - model = init_chat_model( - model="qwen2.5-vl-7b-instruct", # 你的模型名 - model_provider="openai", # 必须填 openai（兼容协议） - base_url="http://127.0.0.1:1234/v1", # 你的自定义 API URL - api_key="1234", # 本地模型随便填 - temperature=0.7, - ) - output = model.invoke("你好").content + model_qwen = create_llm(platform_code='bailian-qwen-turbo') + model_qwen_output = model_qwen.invoke("你好，你是谁").content + print(f"model_qwen output: {model_qwen_output}") - print(f"output:{output}") - return output + model_doubao = create_llm(platform_code='huoshan-doubao') + model_doubao_output = model_doubao.invoke("你好，你是谁").content + print(f"model_doubao output: {model_doubao_output}") + + return { + "model_qwen_output": model_qwen_output, + "model_doubao_output": model_doubao_output + } # Edit this to add the chain you want to add