Merge branch 'master' of https://github.com/Sahandfer/Patient-Simulation

Author: Sahandfer

README.md

@@ -2,7 +2,7 @@
 
 > A unified hub to create, simulate, and evaluate methods for patient/client simulation.
 
-📚 **[Documentation](https://sahandsabour.github.io/PatientHub/)** | 🚀 **[Quick Start](#quick-start)** | 📦 **[Supported Agents](#3-supported-agents)**
+📚 **[Documentation](https://sahandfer.github.io/PatientHub/)** | 🚀 **[Quick Start](#quick-start)** | 📦 **[Supported Agents](#3-supported-agents)**
 
 ## Quick Start
 

data/prompts/therapist/psyche.yaml

@@ -0,0 +1,19 @@
+en: 
+  sys_prompt: |
+    You are an experienced psychiatrist conducting an initial interview with a new patient. Your goal is to gather comprehensive information about the patient’s mental health, symptoms, background, and potential diagnoses. Use a combination of open-ended questions, specific inquiries, and follow-up questions to explore the patient’s situation in depth. Pay attention to both verbal and non-verbal cues, and use your expertise to guide the conversation towards areas that may be particularly relevant for diagnosis.
+
+    The following aspects need to be assessed in the patient: Chief complaint, Present illness, Symptoms, Alleviating factors, Exacerbating factors, Symptom duration, Triggering factors (why the patient decided to visit the hospital today), Stressors, Family history (including diagnoses and substance use), Current family structure, Suicidal ideation, Suicide risk, Self-harming behavior risk, Homicide risk, Suicidal plans, Suicide attempts, Mood, Affect, Verbal productivity, Insight, Perception, Thought process, Thought content, Spontaneity, Social judgment, and Reliability.
+
+    When starting the conversation, begin with exactly these words: "Hello, I’m Dr. Minsoo Kim, what’s your name?". Ask questions one at a time, at most.
+
+    After the interview with the patient is complete, someone will come to ask you about the patient. As an experienced psychiatrist, use appropriate reasoning, your professional judgment, and the information you’ve gathered during the interview to answer their questions. If you cannot determine something even with appropriate reasoning and your expertise, respond with "I don’t know".
+
+zh: 
+  sys_prompt: |
+    你是一名经验丰富的精神科医生，正在对一位新患者进行初诊访谈。你的目标是全面收集患者的心理健康状况、症状、背景以及可能的诊断信息。请结合开放式提问、具体询问与追问，深入探索患者的处境。注意患者的言语与非言语线索，并运用你的专业经验引导对话，聚焦于对诊断可能尤为相关的领域。
+
+    需要评估的方面包括：主诉、现病史、症状、缓解因素、加重因素、症状持续时间、诱发因素（为什么患者今天决定来就诊）、压力源、家族史（包括相关诊断与物质使用）、当前家庭结构、自杀意念、自杀风险、自伤行为风险、他杀风险、自杀计划、自杀尝试、心境、情感、言语产出、自知力、知觉、思维过程、思维内容、自发性、社会判断力以及可靠性。
+
+    开始对话时，请以以下开场白开始："您好，我是金敏洙医生， 请问您叫什么？"。每次最多只问一个问题。
+
+    当你与患者的访谈结束后，会有人来向你询问患者情况。作为一名经验丰富的精神科医生，请运用恰当的推理、你的专业判断以及访谈中收集到的信息来回答他们的问题。如果即使运用恰当的推理与专业经验仍无法确定某件事，请回答："我不知道"。

docs/docs/getting-started/installation.md

@@ -13,8 +13,8 @@ sidebar_position: 1
 
 ```bash
 # Clone the repository
-git clone https://github.com/Sahandfer/Patient-Simulation.git
-cd Patient-Simulation
+git clone https://github.com/Sahandfer/PatientHub.git
+cd PatientHub
 
 # Install dependencies
 uv sync
@@ -27,8 +27,8 @@ source .venv/bin/activate
 
 ```bash
 # Clone the repository
-git clone https://github.com/Sahandfer/Patient-Simulation.git
-cd Patient-Simulation
+git clone https://github.com/Sahandfer/PatientHub.git
+cd PatientHub
 
 # Create virtual environment
 python -m venv .venv
@@ -39,16 +39,44 @@ pip install -e .
 ```
 
 ## Configuration
-
-Create a `.env` file in the project root:
+Create a `.env` file in the project root:  
+### Cloud Models
 
 ```bash
 OPENAI_API_KEY=<your API key>
 OPENAI_BASE_URL=https://api.openai.com
 ```
 
 :::tip Using Other Providers
-PatientHub supports OpenAI, HuggingFace, and local models. See [Configuration](/docs/getting-started/configuration) for details.
+PatientHub supports cloud and local models. See [Configuration](/docs/getting-started/configuration) for details.
+:::
+
+### Local Models via vLLM
+Our recommended setup for local models is to run a vLLM (or OpenAI-compatible) server on a machine/process, then have PatientHub call it over HTTP via `LOCAL_BASE_URL`.
+
+1) Install vLLM on the serving machine (GPU recommended):
+
+```bash
+pip install -U vllm
+```
+
+2) Serve a model with the OpenAI-compatible endpoint (example):
+
+```bash
+vllm serve Qwen/Qwen2.5-7B-Instruct --host 0.0.0.0 --port 8000
+```
+
+3) Point PatientHub to the server in your `.env`:
+
+```bash
+LOCAL_BASE_URL=http://<SERVER_HOST>:8000/v1
+LOCAL_API_KEY=EMPTY
+```
+
+Then set your config to use `model_type=LOCAL` and `model_name` to the model name exposed by your vLLM server.
+
+:::note vLLM fails to start
+it’s usually a CUDA/driver mismatch on the serving machine—check your NVIDIA driver/CUDA runtime and use a vLLM version compatible with your environment.
 :::
 
 ## Verify Installation
@@ -85,10 +113,3 @@ Ensure you've activated the virtual environment:
 ```bash
 source .venv/bin/activate
 ```
-
-**CUDA/GPU issues:**
-For local models, ensure PyTorch is installed with CUDA support:
-
-```bash
-pip install torch --index-url https://download.pytorch.org/whl/cu118
-```

docs/docs/intro.md

@@ -30,6 +30,7 @@ config = OmegaConf.create({
     'temperature': 0.7,
     'max_tokens': 1024,
     'max_retries': 3,
+    'patient_type': 'upset',
     'data_path': 'data/characters/PatientPsi.json',
     'data_idx': 0,
 })

docs/src/pages/index.tsx

@@ -47,8 +47,8 @@ function QuickStart() {
             <pre className={styles.codeBlock}>
               <code>
                 {`# Clone the repository
-git clone https://github.com/Sahandfer/Patient-Simulation.git
-cd Patient-Simulation
+git clone https://github.com/Sahandfer/PatientHub.git
+cd PatientHub
 
 # Install dependencies
 uv sync

patienthub/therapists/__init__.py

@@ -2,6 +2,8 @@
 from .user import UserTherapist, UserTherapistConfig
 from .CBT import CBTTherapist, CBTTherapistConfig
 from .bad import BadTherapist, BadTherapistConfig
+from .psyche import PsycheTherapist, PsycheTherapistConfig
+
 
 
 from omegaconf import DictConfig
@@ -12,6 +14,7 @@
     "user": UserTherapist,
     "CBT": CBTTherapist,
     "bad": BadTherapist,
+    "psyche": PsycheTherapist,
 }
 
 # Registry of therapist configs (for Hydra registration)
@@ -20,6 +23,7 @@
     "user": UserTherapistConfig,
     "CBT": CBTTherapistConfig,
     "bad": BadTherapistConfig,
+    "psyche": PsycheTherapistConfig,
 }
 
 

patienthub/therapists/psyche.py

@@ -0,0 +1,43 @@
+from typing import Dict, List
+from omegaconf import DictConfig
+from dataclasses import dataclass
+
+from patienthub.base import ChatAgent
+from patienthub.configs import APIModelConfig
+from patienthub.utils import load_prompts, load_json, get_chat_model
+
+@dataclass
+class PsycheTherapistConfig(APIModelConfig):
+    """
+    Configuration for the PsycheTherapist agent.
+    """
+    agent_type: str = "psyche"
+
+class PsycheTherapist(ChatAgent):
+    def __init__(self, configs: DictConfig):
+        self.configs = configs
+
+        self.name = "Dr. Minsoo Kim"
+
+        self.chat_model = get_chat_model(configs)
+        self.prompts = load_prompts(
+            role="therapist", agent_type="psyche", lang=configs.lang
+        )
+        self.build_sys_prompt()
+
+    def build_sys_prompt(self):
+        self.messages = [{"role": "system", "content": self.prompts["sys_prompt"].render()}]
+
+    def set_client(self, client, prev_sessions: List[Dict[str, str] | None] = []):
+        self.client = client.get("name", "client")
+
+    def generate_response(self, msg: str):
+        self.messages.append({"role": "user", "content": msg})
+        res = self.chat_model.generate(self.messages)
+        self.messages.append({"role": "assistant", "content": res.content})
+
+        return res
+
+    def reset(self):
+        self.build_sys_prompt()
+        self.client = None

pyproject.toml

@@ -24,4 +24,8 @@ dev = [
     "jupyter>=1.1.1",
     "ipykernel>=7.0.1",    
     "chainlit>=2.8.3",
-]
+]
+
+[tool.setuptools.packages.find]
+include = ["patienthub*"] 
+exclude = ["data*"]