virtual-clinic

Python client for the Virtual Clinic REST API — multi-turn conversations with LLM-based simulated patient agents powered by Synthea-generated electronic health records.

Installation

pip install virtual-clinic

Or install from source (for development):

# From the repo root
cd packages/client
uv pip install -e .

Quick Start

from virtual_clinic import VirtualClinic

client = VirtualClinic(
    base_url="https://virtual-clinic-api.vercel.app",
    token="your-jwt-token",  # provided by workshop organizers
)

# 1. List available patients (requires admin token)
patients = client.patients.list(page=1, limit=10)
for p in patients.data:
    print(f"{p.first} {p.last} — {p.gender}, born {p.birth_date}")

print(f"Page {patients.pagination.page} of {patients.pagination.total_pages}")

# 2. Inspect a patient's full EHR
detail = client.patients.get(patients.data[0].id)
print(f"Active conditions: {detail.summary.active_conditions}")
print(f"Active medications: {detail.summary.active_medications}")
print(f"Allergies: {detail.summary.allergies}")

# 3. Start a diagnostic conversation
convo = client.conversations.create(
    patient_id=patients.data[0].id,
    task_type="diagnosis",
)
print(f"Conversation {convo.id} started with {convo.patient_name}")

# 4. Interview the simulated patient
reply = client.conversations.send_message(
    convo.id,
    content="Hello, what brings you in today?",
)
print(f"Patient: {reply.content}")

# Continue the conversation...
reply = client.conversations.send_message(
    convo.id,
    content="How long have you been experiencing these symptoms?",
)
print(f"Patient: {reply.content}")

# 5. Review the full conversation history
history = client.conversations.get(convo.id)
for msg in history.messages:
    print(f"[{msg.role}] {msg.content[:80]}...")

# 6. Clean up
client.close()

Context Manager

The client supports the context manager protocol for automatic cleanup:

with VirtualClinic(base_url="...", token="...") as client:
    health = client.health()
    print(health.status, health.database)

API Reference

VirtualClinic(*, base_url, token, timeout=60.0)

The main client. All parameters are keyword-only.

Health

client.health() -> HealthStatus

Check API health (public, no auth required).

Patients (admin token required)

client.patients.list(*, page=1, limit=20) -> PaginatedResponse[PatientSummary]
client.patients.get(patient_id: str) -> PatientDetail

Conversations

client.conversations.list(*, page=1, limit=20, patient_id=None, task_type=None) -> PaginatedResponse[ConversationSummary]
client.conversations.create(*, patient_id, task_type, metadata=None) -> CreatedConversation
client.conversations.get(conversation_id: str) -> ConversationWithMessages
client.conversations.send_message(conversation_id: str, *, content: str) -> AssistantMessage

Task Types

Error Handling

All API errors raise typed exceptions:

from virtual_clinic import (
    VirtualClinicError,     # base class — catch-all
    AuthenticationError,    # 401 — invalid or missing token
    ForbiddenError,         # 403 — insufficient permissions
    NotFoundError,          # 404 — resource not found
    ValidationError,        # 400 — invalid request
    ServerError,            # 5xx — server error
    ConnectionError,        # network/DNS/timeout failure
)

try:
    patient = client.patients.get("nonexistent-uuid")
except NotFoundError as e:
    print(f"Patient not found: {e.message}")
except AuthenticationError:
    print("Check your token!")
except VirtualClinicError as e:
    print(f"Something went wrong: {e.message}")

API error exceptions expose these attributes:

Models

All API responses are returned as Pydantic v2 models with full type hints. This gives you autocomplete in IDEs, runtime validation, and easy serialization:

# Convert to dict
patient_dict = patient.model_dump()

# Convert to JSON string
patient_json = patient.model_dump_json()

# Access fields with autocomplete
print(detail.summary.active_conditions)
print(detail.patient.first, detail.patient.last)

Requirements

• Python >= 3.10
• httpx >= 0.27
• pydantic >= 2.0