File size: 6,421 Bytes
82bf89e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 |
from asgi_correlation_id import correlation_id
from fastapi import APIRouter
from fastapi.responses import StreamingResponse, JSONResponse
from utils.bio_logger import bio_logger as logger
from utils.i18n_util import (
get_language,
create_error_response,
)
from utils.i18n_context import with_language
from bio_requests.chat_request import ChatRequest
from service.chat import ChatService
router = APIRouter(prefix="/mcp", tags=["MCP"])
@router.post("/bio_qa", response_model=None, operation_id="bio_qa_stream_chat")
async def bio_qa(query: str, lang: str = "en"):
"""
Biomedical Q&A with Evidence-Based RAG System
This MCP tool provides comprehensive, research-backed answers to biological and medical questions
using a sophisticated Retrieval-Augmented Generation (RAG) system. The tool performs multi-source
research and evidence-based synthesis to deliver accurate, well-cited responses.
**Process Overview:**
1. **Query Analysis & Rewriting** (30-45 seconds)
- Analyzes the user's question and identifies key biomedical concepts
- Performs intelligent query rewriting to improve search effectiveness
- Generates multiple search variations to capture different aspects of the question
- Optimizes search terms for both PubMed and web search engines
2. **Multi-Source Literature Search** (60-90 seconds)
- **PubMed Database Search**: Searches scientific literature database for peer-reviewed papers
- **Web Search**: Conducts web searches for recent developments, clinical guidelines, and additional context
- **Concurrent Processing**: Performs both searches simultaneously for efficiency
- **Content Extraction**: Extracts and processes relevant content from search results
3. **Intelligent Reranking** (30-45 seconds)
- Ranks search results by relevance to the specific question
- Filters out low-quality or irrelevant content
- Prioritizes recent, authoritative, and highly relevant sources
- Ensures diversity in source types (papers, guidelines, reviews, etc.)
4. **Evidence-Based Answer Generation** (60-90 seconds)
- Synthesizes information from multiple high-quality sources
- Generates comprehensive, well-structured answers
- Includes proper citations and references
- Provides evidence-based explanations with source attribution
**Input:**
- query (string): A biological or medical question
Examples: "What causes Alzheimer's disease?", "How do mRNA vaccines work?",
"What are the latest treatments for diabetes?", "Explain CRISPR gene editing"
- lang (string, optional): Language preference ("en" for English, "zh" for Chinese)
- is_pubmed (boolean, optional): Enable PubMed scientific literature search (default: True)
- When True: Searches peer-reviewed scientific papers for authoritative evidence
- When False: Skips PubMed search to reduce processing time
- is_web (boolean, optional): Enable web search for additional context (default: True)
- When True: Searches web for recent developments, clinical guidelines, and additional context
- When False: Skips web search to reduce processing time
**Output:**
- A comprehensive answer with the following components:
* **Main Answer**: Evidence-based response to the question
* **Citations**: Properly formatted references to source materials
* **Source Links**: Direct links to PubMed papers and web sources
* **Evidence Summary**: Overview of the evidence supporting the answer
**Key Features:**
- **Real-time Streaming**: Provides progress updates via Server-Sent Events (SSE)
- **Multi-Source Research**: Combines PubMed scientific literature with web-based information
- **Intelligent Query Processing**: Uses advanced query rewriting for better search results
- **Quality Control**: Reranks results to ensure relevance and authority
- **Evidence-Based Answers**: All claims are supported by cited sources
- **Comprehensive Coverage**: Covers genetics, molecular biology, diseases, treatments, and more
**Expected Duration:** 3 minutes (may vary based on query complexity and search configuration)
**Performance Notes:**
- Full search (is_pubmed=True, is_web=True): ~3 minutes with comprehensive coverage
- PubMed only (is_pubmed=True, is_web=False): ~2 minutes, focused on scientific literature
- Web only (is_pubmed=False, is_web=True): ~2 minutes, focused on recent developments
- Minimal search (is_pubmed=False, is_web=False): ~1 minute, basic query processing only
**Use Cases:**
- Medical education and learning
- Clinical decision support
- Research background information
- Patient education content
- Healthcare professional training
- Scientific literature exploration
**Evidence Quality:**
- Primary sources from peer-reviewed scientific journals
- Recent clinical guidelines and recommendations
- Authoritative medical websites and databases
- Multiple source verification for key claims
**Note:** This tool is specifically optimized for biomedical and healthcare questions.
For best results, provide specific, well-defined questions about biological or medical topics.
"""
logger.info(f"{correlation_id.get()} Bio QA for {query}")
chat_request = ChatRequest(query=query, language=lang, is_pubmed=True, is_web=True)
# 解析语言设置
language = get_language(chat_request.language)
# 使用上下文管理器设置语言
with with_language(language):
try:
chat_service = ChatService()
return StreamingResponse(
chat_service.generate_stream(chat_request),
media_type="text/event-stream",
headers={
"Connection": "keep-alive",
"Cache-Control": "no-cache",
},
)
except Exception as e:
logger.error(f"{correlation_id.get()} Stream chat error: {e}")
error_response = create_error_response(
error_key="service_unavailable",
details=str(e),
error_code=500,
)
return JSONResponse(content=error_response, status_code=500)
|