电子健康记录AI接口服务

Blood Pressure (Archetype)
├── Systolic: 120 mmHg
├── Diastolic: 80 mmHg
├── Position: Sitting
└── Time: 2025-12-14 10:30:00

Query: "patient has persistent cough and fever"
Embedding: [0.12, 0.45, -0.23, ...]

Qdrant finds similar vectors:
✓ R05 - Cough (similarity: 91.2%)
✓ R50.9 - Fever, unspecified (similarity: 88.7%)
✓ J00 - Acute nasopharyngitis (similarity: 85.3%)

from fastmcp import FastMCP

mcp = FastMCP("openEHR")

@mcp.tool()
def openehr_ehr_create(subject_id: str) -> dict:
    """Create a new Electronic Health Record"""
    # Implementation connects to EHRbase API
    return {"ehr_id": "...", "subject_id": subject_id}

AI Request → MCP Server → EHRbase API → PostgreSQL Database
              ↓
         Validates against
         openEHR Templates

Clinical Text: "Patient with chronic cough and fever"
        ↓
  Sentence Transformer (all-mpnet-base-v2)
        ↓
  Vector Embedding [768 dimensions]
        ↓
  Qdrant Semantic Search
        ↓
Results:
  • R05 - Cough (91.2% match)
  • R50.9 - Fever (88.7% match)

┌─────────────────────────────────────────────────────────────┐
│                     Claude / AI Assistant                   │
└────────────────────────┬────────────────────────────────────┘
                         │ MCP Protocol
                         │
┌────────────────────────▼────────────────────────────────────┐
│                  openEHR MCP Server (FastMCP)               │
│  ┌──────────────┐  ┌──────────────┐ ┌───────────────┐       │
│  │   EHR Tools  │  │ Query Tools  │ │Medical Coding │       │
│  └──────┬───────┘  └───────┬──────┘ └─────────┬─────┘       │
└─────────┼──────────────────┼──────────────────┼─────────────┘
          │                  │                  │
          ▼                  ▼                  ▼
  ┌───────────────┐  ┌───────────────┐  ┌──────────────┐
  │   EHRbase     │  │  PostgreSQL   │  │   Qdrant     │
  │ (openEHR API) │  │   Database    │  │  Vector DB   │
  └───────────────┘  └───────────────┘  └──────────────┘

# Clone the repository
git clone https://github.com/yourusername/openehr-mcp-server.git
cd openehr-mcp-server

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Copy example environment file
cp .env.example .env

# Edit .env and add your API key
# GEMINI_API_KEY=your_actual_api_key_here

# Start EHRbase and Qdrant
cd docker-compose
docker compose up -d

# Verify services are running
docker compose ps

# Copy the example config
cp mcp-config.example.json mcp-config.json

# Edit with your actual project paths
# Replace /path/to/your/project with your installation directory

{
  "mcpServers": {
    "openEHR": {
      "command": "/path/to/your/project/venv/bin/python",
      "args": ["/path/to/your/project/src/openehr_mcp_server.py"],
      "env": {
        "EHRBASE_URL": "http://localhost:8080/ehrbase/rest",
        "EHRBASE_JSON_FORMAT": "wt_flat",
        "PYTHONPATH": "/path/to/your/project/src"
      }
    }
  }
}

# Upload the default template
python scripts/upload_template.py

# Or specify a custom template
python scripts/upload_template.py --template ehr-templates/patient_visit_template.opt

# Create EHR with random subject ID
python scripts/create_ehr.py

# Or create with specific patient identifier
python scripts/create_ehr.py --subject-id "MRN-12345"

# 1. Ensure Qdrant is running
curl http://localhost:6333/healthz

# 2. Prepare ICD-10 dataset (diagnosis.csv)
# Place your CSV file in the project root directory or specify path with --csv
# Required columns: code, short, long
# Example sources: WHO ICD-10 catalog, CMS, or other medical coding databases

# 3. Load embeddings into Qdrant (from project root)
python scripts/embedding.py

# Or specify custom CSV location
python scripts/embedding.py --csv /path/to/your/diagnosis.csv

# Use a different embedding model
python scripts/embedding.py --model sentence-transformers/all-MiniLM-L6-v2

# Check if collection exists
curl http://localhost:6333/collections

# Check collection size
curl http://localhost:6333/collections/icd_mpnet_basev2

# Test ICD-10 code suggestion
python tests/test_icd_coding.py

# Test with MCP protocol
python tests/test_mcp_cholera.py

# If embeddings are not loaded, these tests will fail with:
# "Collection not found" or "No results returned"
## 🧪 Testing & Validation

### Run Integration Tests

```bash
# Install test dependencies
pip install -r requirements-test.txt

# Run all tests
pytest tests/ -v

# Run specific test file
pytest tests/test_ehr_client.py -v

# Run with coverage
pytest tests/ --cov=src --cov-report=html

# Test ICD-10 code suggestion
python tests/test_icd_coding.py

# Test with MCP protocol
python tests/test_mcp_cholera.py

# Check Docker services
docker compose ps

# Test EHRbase
curl http://localhost:8080/ehrbase/rest/openehr/v1/definition/template/adl1.4

# Test Qdrant
curl http://localhost:6333/healthz

# View service logs
docker compose logs -f ehrbase
docker compose logs -f qdrant

# Install streamlit if not already installed
pip install streamlit

# Launch the web UI
streamlit run app.py

openehr-mcp-server/
├── src/                          # Source code
│   ├── openehr_mcp_server.py    # Main MCP server
│   ├── medical_coding.py        # ICD-10 coding logic
│   ├── mcp_prompts.py          # MCP prompt definitions
│   └── ehrbase/                # EHRbase API clients
├── scripts/                    # Utility scripts
│   ├── create_ehr.py          # EHR creation
│   ├── upload_template.py     # Template upload
│   ├── embedding.py           # ICD-10 embedding generator
│   └── vector_search.py       # Vector search utilities
├── tests/                       # Test suite
├── docker-compose/             # Docker services configuration
│   └── docker-compose.yml      # EHRbase + PostgreSQL + Qdrant
├── ehr-templates/              # openEHR templates (.opt files)
├── DOCS/                       # Documentation
│   ├── SETUP.md               # Detailed setup guide
│   └── MCP_TEST_GUIDE.md      # Testing guide
├── app.py                      # Streamlit Web UI for EHRbase (optional)
├── diagnosis.csv               # ICD-10 dataset (place here or specify path)
├── .env.example               # Environment template
├── mcp-config.example.json    # MCP configuration template
├── requirements.txt           # Python dependencies
└── README.md                  # This file

# Ensure EHRbase is running
cd docker-compose && docker compose up -d

# Launch Streamlit
streamlit run app.py

I need to record vital signs for patient EHR ID: abc-123-xyz

Blood Pressure: 120/80 mmHg
Heart Rate: 72 bpm
Temperature: 98.6°F
Respiratory Rate: 16 breaths/min

Show me all vital signs recordings for the last 7 days
for EHR ID: abc-123-xyz

Suggest ICD-10 codes for: "Patient presents with persistent 
dry cough, mild fever, and shortness of breath"

Found 3 ICD-10 codes:

1. R05 - Cough (91.2% match)
2. R50.9 - Fever, unspecified (88.7% match)
3. R06.02 - Shortness of breath (86.4% match)
**In Claude Desktop:**


Suggest ICD-10 codes for: "Patient presents with persistent 
dry cough, mild fever, and shortness of breath"

# Check if services are running
docker compose ps

# 1. Ensure Qdrant is running
curl http://localhost:6333/healthz

# 2. Check if collection exists
curl http://localhost:6333/collections

# 3. If collection is missing, load embeddings
python scripts/embedding.py

# 4. Verify collection was created
curl http://localhost:6333/collections/icd_all_mpnet_base_v2

# Upload the template
python scripts/upload_template.py

# Verify upload
curl http://localhost:8080/ehrbase/rest/openehr/v1/definition/template/adl1.4

# Ensure Qdrant is running
curl http://localhost:6333/healthz

# Check if collection exists
curl http://localhost:6333/collections