Merge pull request #172 from gomate-community/pipeline

Pipeline

Author: yanqiangmiffy

Dockerfile

@@ -1,5 +1,5 @@
 # Use the official Ubuntu base image
-FROM pytorch/pytorch:2.6.0-cuda12.6-cudnn9-runtime
+FROM pytorch/pytorch:2.6.0-cuda12.6-cudnn9-devel
 ENV DEBIAN_FRONTEND=noninteractive
 ENV CUDA_DEVICE_ORDER=PCI_BUS_ID
 ENV PYTORCH_NVML_BASED_CUDA_CHECK=1

api/README.md

@@ -0,0 +1,24 @@
+# 使用默认 .env，再覆盖 RERANKER_NAME
+sh start_rag.sh -e RERANKER_NAME=G:/pretrained_models/mteb/bge-reranker-large-v2
+
+
+docker run --rm \
+  --env-file /path/to/.env \
+  -e RERANKER_NAME=G:/pretrained_models/mteb/bge-reranker-large-v2 \
+  -p 10000:10000 \
+  -v /g/Projects/TrustRAG:/app \
+  -w /app/api/rag \
+  trustrag:v0.1 \
+  sh -c "PYTHONPATH=/app python main.py"
+
+
+
+docker run --rm \
+  --gpus all \
+  -p 10000:10000 \
+  -v /mnt/g/pretrained_models/mteb/bge-reranker-large:/mnt/g/pretrained_models/mteb/bge-reranker-large \
+  -v .:/app \
+  -e RERANKER_NAME=/mnt/g/pretrained_models/mteb/bge-reranker-large \
+  -w /app/api/rag \
+  trustrag:v0.1 \
+  sh -c "PYTHONPATH=/app python main.py"

api/rag/.env.example

@@ -0,0 +1,32 @@
+### API server
+API_HOST=0.0.0.0
+API_PORT=10000
+API_URL=http://127.0.0.1:10001
+
+### Upstream LLM service (gomall)
+GOMALL_BASE_URL=http://10.208.61.1:32004/api/v1/1504_gomall_qwen3/Qwen3-30B-A3B-Instruct-2507/
+GOMALL_API_KEY=sk-xx
+
+### Local rerank service
+RERANK_BASE_URL=
+RERANK_API_KEY=
+
+### Model configs
+LLM_NAME=Qwen3-30B-A3B-Instruct-2507
+LLM_PATH=G:/pretrained_models/llm/glm-4-9b-chat
+LLM_SERVICE=online
+
+EMBEDDING_NAME=G:/pretrained_models/mteb/bge-large-zh-v1.5
+EMBEDDING_PATH=G:/pretrained_models/mteb/bge-large-zh-v1.5
+EMBEDDING_SERVICE=local
+
+RERANKER_NAME=G:/pretrained_models/mteb/bge-reranker-large
+RERANKER_PATH=G:/pretrained_models/mteb/bge-reranker-large
+RERANKER_SERVICE=local
+
+### Paths
+DOCS_PATH=G:/Projects/TrustRAG/data/docs
+INDEX_PATH=G:/Projects/TrustRAG/examples/retrievers/dense_cache
+
+### Rewriter tool
+REWRITER_API_URL=http://10.208.61.1:32004/api/v1/1504_gomall_qwen3/Qwen3-30B-A3B-Instruct-2507/

api/rag/apps/config/__init__.py

@@ -1,11 +1,3 @@
-#!/usr/bin/env python
-# -*- coding:utf-8 _*-
-"""
-@author:quincy qiang
-@license: Apache Licence
-@file: __init__.py
-@time: 2024/06/13
-@contact: [email protected]
-@software: PyCharm
-@description: coding..
-"""
+from .base_config import RAGSettings
+
+settings = RAGSettings()

api/rag/apps/config/app_config.py

@@ -1,20 +1,9 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 _*-
 """
-@author:quincy qiang
-@license: Apache Licence
-@file: app_config.py.py
-@time: 2024/06/13
-@contact: [email protected]
-@software: PyCharm
-@description: coding..
+应用基础配置。
 """
-import pprint
-from typing import ClassVar
-
-pp = pprint.PrettyPrinter(indent=4)
-
-
+from api.rag.apps.config import settings
 class AppConfig:
     """配置类"""
     API_V1_STR: str = ""
@@ -37,13 +26,13 @@ class AppConfig:
         {"url": "/v2", "description": "测试地址"},
     ]
 
-    WEB_URL: ClassVar[str] = '*'
+    WEB_URL: str = '*'
     # 接口地址
-    API_URL: ClassVar[str] = 'http://127.0.0.1:10001'
+    API_URL: str = settings.api_url
     # 运行访问的地址
-    API_HOST: ClassVar[str] = '0.0.0.0'
+    API_HOST: str = settings.api_host
     # 端口
-    API_PORT: int = 10000
+    API_PORT: int = settings.api_port
 
     DEBUGGER: bool = True
 

api/rag/apps/config/base_config.py

@@ -0,0 +1,40 @@
+# from pydantic import BaseSettings
+from pydantic import Field
+from pydantic_settings import BaseSettings, SettingsConfigDict
+
+
+class RAGSettings(BaseSettings):
+    """Centralized application settings loaded from environment variables."""
+
+    # API service
+    api_host: str = Field(default="0.0.0.0", validation_alias="API_HOST")
+    api_port: int = Field(default=10000, validation_alias="API_PORT")
+    api_url: str = Field(default="http://127.0.0.1:10001", validation_alias="API_URL")
+
+    # Upstream LLM service (gomall)
+    gomall_base_url: str = Field(default="http://10.208.61.1:32004/api/v1/1504_gomall_qwen3/Qwen3-30B-A3B-Instruct-2507/", validation_alias="GOMALL_BASE_URL")
+    gomall_api_key: str = Field(default="", validation_alias="GOMALL_API_KEY")
+    llm_name: str = Field(default="Qwen3-30B-A3B-Instruct-2507", validation_alias="LLM_NAME")
+    # Tool specific
+    rewriter_api_url: str = Field(default="http://10.208.61.1:32004/api/v1/1504_gomall_qwen3/Qwen3-30B-A3B-Instruct-2507/", validation_alias="REWRITER_API_URL")
+
+    # Local rerank service
+    rerank_base_url: str | None = Field(default=None, validation_alias="RERANK_BASE_URL")
+    rerank_api_key: str | None = Field(default=None, validation_alias="RERANK_API_KEY")
+
+
+    embedding_name: str = Field(default="G:/pretrained_models/mteb/bge-large-zh-v1.5", validation_alias="EMBEDDING_NAME")
+    embedding_path: str = Field(default="G:/pretrained_models/mteb/bge-large-zh-v1.5", validation_alias="EMBEDDING_PATH")
+
+    reranker_name: str = Field(default="G:/pretrained_models/mteb/bge-reranker-large", validation_alias="RERANKER_NAME")
+    reranker_path: str = Field(default="G:/pretrained_models/mteb/bge-reranker-large", validation_alias="RERANKER_PATH")
+    # Paths
+    docs_path: str = Field(default="G:/Projects/TrustRAG/data/docs", validation_alias="DOCS_PATH")
+    index_path: str = Field(default="G:/Projects/TrustRAG/examples/retrievers/dense_cache", validation_alias="INDEX_PATH")
+
+    model_config = SettingsConfigDict(
+        env_file=".env",
+        env_file_encoding="utf-8",
+        case_sensitive=False,
+        extra="ignore",
+    )

api/rag/apps/config/rerank_config.py

@@ -9,17 +9,11 @@
 @software: PyCharm
 @description: coding..
 """
-from trustrag.config.config_loader import config
+from api.rag.apps.config import settings
 
-class RerankConfig():
+class RerankConfig:
     """重排序配置类"""
-    # 从配置文件加载服务和模型配置
-    _rerank_service = config.get_config('services.rerank')
-    _rerank_model = config.get_config('models.reranker')
-    
-    # 模型名称
-    model_name_or_path:str = _rerank_model['name']
-    # 服务 URL
-    base_url:str = _rerank_service['base_url']
-    # API 密钥
-    api_key:str = _rerank_service['api_key']
+
+    model_name_or_path: str = settings.reranker_name
+    base_url: str | None = settings.rerank_base_url
+    api_key: str | None = settings.rerank_api_key

api/rag/apps/core/judge/views.py

@@ -11,30 +11,26 @@
 """
 import loguru
 from fastapi import APIRouter
-from trustrag.config.config_loader import config
+
+from api.rag.apps.config import settings
 from api.rag.apps.core.judge.bodys import JudgeBody
 from api.rag.apps.handle.response.json_response import ApiResponse
 from trustrag.modules.judger.bge_judger import BgeJudger, BgeJudgerConfig
 from trustrag.modules.judger.chatgpt_judger import OpenaiJudger, OpenaiJudgerConfig
 
 judge_router = APIRouter()
 
-# 加载服务和模型配置
-llm_service = config.get_config('services.dmx')
-llm_model = config.get_config('models.llm')
-rerank_model = config.get_config('models.reranker')
-
 # BGE 判断器配置
 judge_config = BgeJudgerConfig(
-    model_name_or_path=rerank_model['name']
+    model_name_or_path=settings.reranker_name,
 )
 bge_judger = BgeJudger(judge_config)
 
 # LLM 判断器配置
 judger_config = OpenaiJudgerConfig(
-    base_url=llm_service['base_url'],
-    api_key=llm_service['api_key'],
-    model_name=llm_model['name']
+    base_url=settings.gomall_base_url,
+    api_key=settings.gomall_api_key,
+    model_name=settings.llm_name,
 )
 openai_judger = OpenaiJudger(judger_config)
 

api/rag/apps/core/rerank/views.py

@@ -12,25 +12,18 @@
 import loguru
 from fastapi import APIRouter
 
+from api.rag.apps.config import settings
 from api.rag.apps.core.rerank.bodys import RerankBody
 from api.rag.apps.core.rerank.models import Application
 from api.rag.apps.handle.response.json_response import UserNotFoundResponse, ApiResponse
 from trustrag.modules.reranker.bge_reranker import BgeReranker, BgeRerankerConfig
-from trustrag.config.config_loader import config
-
-# from apps.handle.exception.exception import MallException
-# from apps.core.config.models import LLMModel
-# from tortoise.contrib.pydantic import pydantic_model_creator
 
 rerank_router = APIRouter()
-# 从配置文件加载重排序配置
-rerank_service = config.get_config('services.rerank')
-rerank_model = config.get_config('models.reranker')
 
 reranker_config = BgeRerankerConfig(
-    model_name_or_path=rerank_model['name'],
-    api_key=rerank_service['api_key'],
-    url=rerank_service['base_url']
+    model_name_or_path=settings.reranker_name,
+    api_key=settings.rerank_api_key,
+    url=settings.rerank_base_url,
 )
 bge_reranker = BgeReranker(reranker_config)
 # Create

api/rag/apps/core/rewrite/views.py

@@ -12,15 +12,18 @@
 import loguru
 from fastapi import APIRouter
 
+from api.rag.apps.config import settings
 from api.rag.apps.core.rewrite.bodys import RewriteBody
 from api.rag.apps.handle.response.json_response import ApiResponse
-from trustrag.modules.rewriter.openai_rewrite import OpenaiRewriter,OpenaiRewriterConfig
+from trustrag.modules.rewriter.openai_rewrite import OpenaiRewriter, OpenaiRewriterConfig
 
 
 rewriter_router = APIRouter()
 
 rewriter_config = OpenaiRewriterConfig(
-    api_url="http://10.208.63.29:8888"
+    base_url=settings.rewriter_api_url,
+    api_key=settings.gomall_api_key,
+    model_name=settings.llm_name,
 )
 openai_rewriter = OpenaiRewriter(rewriter_config)