# ============================================================
# Installation for GPU llama-cpp-python
# ============================================================
!CMAKE_ARGS="-DLLAMA_CUBLAS=on" FORCE_CMAKE=1 pip install llama-cpp-python==0.2.28 --force-reinstall --no-cache-dir -q 2>/dev/null

# Installing other dependencies
!pip install huggingface_hub==0.35.3 pandas==2.2.2 tiktoken==0.12.0 pymupdf==1.26.5 langchain==0.3.27 langchain-community==0.3.31 chromadb==1.1.1 sentence-transformers==5.1.1 numpy==1.26.4 -q 2>/dev/null

     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.0/9.4 MB ? eta -:--:--
     ━╺━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.3/9.4 MB 9.1 MB/s eta 0:00:01
     ━━━━━━━━━━╺━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.4/9.4 MB 35.6 MB/s eta 0:00:01
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 9.4/9.4 MB 95.0 MB/s eta 0:00:00
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Installing backend dependencies ... done
  Preparing metadata (pyproject.toml) ... done
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 45.5/45.5 kB 292.1 MB/s eta 0:00:00
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16.4/16.4 MB 278.7 MB/s eta 0:00:00
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 44.6/44.6 kB 244.2 MB/s eta 0:00:00
  Building wheel for llama-cpp-python (pyproject.toml) ... done

# ============================================================
# Core Imports and Shared Configuration
# ============================================================

# Dependencies
import os
import pandas as pd
from huggingface_hub import hf_hub_download
from llama_cpp import Llama
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.document_loaders import PyMuPDFLoader
from langchain_community.embeddings.sentence_transformer import SentenceTransformerEmbeddings
from langchain_community.vectorstores import Chroma

# Global Qs
questions = [
    "What are the common symptoms and treatments for pulmonary embolism?",
    "What are the common symptoms for appendicitis, and can it be cured via medicine? If not, what surgical procedure should be followed to treat it?",
    "Can you provide the trade names of medications used for treating hypertension?",
    "What are the first-line options and alternatives for managing rheumatoid arthritis?",
    "What are the effective treatments or solutions for addressing sudden patchy hair loss, commonly seen as localized bald spots on the scalp, and what could be the possible causes behind it?",
    "What treatments are recommended for a person who has sustained a physical injury to brain tissue, resulting in temporary or permanent impairment of brain function?",
    "What are the necessary precautions and treatment steps for a person who has fractured their leg during a hiking trip, and what should be considered for their care and recovery?",
    "What are the diagnostic steps for suspected endocrine disorders?",
    "What is the protocol for managing sepsis in a critical care unit?"
]

# ============================================================
# Load and Initialize LLM
# ============================================================


# Load LLM
model_name_or_path = "TheBloke/Mistral-7B-Instruct-v0.2-GGUF"
model_basename = "mistral-7b-instruct-v0.2.Q4_K_M.gguf"

model_path = hf_hub_download(
    repo_id=model_name_or_path,
    filename=model_basename
)

# Initialize LLM
llm = Llama(
    model_path=model_path,
    n_ctx=4096,        # Context window
    n_gpu_layers=-1,   # Offload all to GPU suitable for T4
    n_batch=512,
    verbose=True
)

/usr/local/lib/python3.12/dist-packages/huggingface_hub/utils/_auth.py:94: UserWarning: 
The secret `HF_TOKEN` does not exist in your Colab secrets.
To authenticate with the Hugging Face Hub, create a token in your settings tab (https://huggingface.co/settings/tokens), set it as secret in your Google Colab and restart your session.
You will be able to reuse this secret in all of your notebooks.
Please note that authentication is recommended but still optional to access public models or datasets.
  warnings.warn(

mistral-7b-instruct-v0.2.Q4_K_M.gguf:   0%|          | 0.00/4.37G [00:00<?, ?B/s]

AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 |

# ============================================================
# Inference Wrapper
# ============================================================

def generate_response(query, max_tokens=2048, temperature=0.1, top_p=0.95, top_k=50, system_prompt=None):
    if system_prompt:
        prompt = f"[INST] {system_prompt}\n{query} [/INST]"
    else:
        prompt = f"[INST] {query} [/INST]"

    model_output = llm(
        prompt=prompt,
        max_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        top_k=top_k
    )
    return model_output['choices'][0]['text'].strip()

# ============================================================
# Baseline Question Execution
# ============================================================

vanilla_results = []
for i, q in enumerate(questions, 1):
    print(f"--- Question {i} ---")
    print(f"Q: {q}")
    ans = generate_response(q)
    print(f"A: {ans}\n")
    vanilla_results.append(ans)

--- Question 1 ---
Q: What are the common symptoms and treatments for pulmonary embolism?
A: A pulmonary embolism (PE) is a serious condition that occurs when a blood clot or other foreign substance travels to the lungs and blocks one or more of the arteries supplying the lungs with blood. This can lead to decreased oxygen supply to the body, which can be life-threatening. Here are some common symptoms of PE:

1. Shortness of breath: This is often the first symptom of a PE and may worsen with exertion or lying flat in bed.
2. Chest pain: This can be described as sharp, stabbing, or crushing pain that worsens with deep breathing or coughing.
3. Rapid heart rate: A fast heartbeat (tachycardia) is a common symptom of PE and may be present at rest or with exertion.
4. Cough: Some people with PE may have a persistent cough, which may bring up blood-tinged sputum.
5. Swelling: Swelling in the legs, ankles, or feet can occur due to the clot obstructing blood flow in the veins of the lower extremities.
6. Lightheadedness or fainting: In severe cases, PE can cause dizziness, weakness, or even loss of consciousness.

Treatment for PE depends on the severity of the condition and may include:

1. Anticoagulation therapy: This involves administering medications to prevent blood clots from forming or growing larger. These medications are typically given intravenously in the hospital and then continued as an oral medication at home.
2. Thrombolysis: In some cases, clot-dissolving medications (thrombolytics) may be used to quickly restore blood flow to the lungs. This is usually reserved for people with severe PE or those who are not responding well to anticoagulation therapy.
3. Embolectomy: This is a surgical procedure in which the clot is removed from the pulmonary artery. It may be recommended for people with large, life-threatening clots that are not responding to other treatments.
4. Oxygen therapy: People with PE may require oxygen therapy to help improve their breathing and increase oxygen levels in the body.
5. Supportive care: This may include bed rest, fluids, and medications to manage symptoms such as pain, shortness of breath, and heart rate.

It is important to note that PE can be a life-threatening condition, so if you suspect you or someone else may have this condition, seek medical help immediately. Early diagnosis and treatment are crucial for improving outcomes and reducing the risk of complications.

--- Question 2 ---
Q: What are the common symptoms for appendicitis, and can it be cured via medicine? If not, what surgical procedure should be followed to treat it?

Llama.generate: prefix-match hit

A: Appendicitis is a medical condition characterized by inflammation of the appendix, a small tube-shaped organ located in the lower right side of the abdomen. The following are common symptoms of appendicitis:

1. Abdominal pain: The pain is typically located in the lower right quadrant of the abdomen and may begin as a mild discomfort that gradually worsens over time. It may be constant or come and go, and it may be accompanied by cramping or bloating.
2. Loss of appetite: A loss of interest in food and a feeling of nausea are common symptoms of appendicitis.
3. Nausea and vomiting: Vomiting is a common symptom of appendicitis, especially as the condition progresses.
4. Fever: A fever may be present, particularly if the appendicitis has advanced.
5. Diarrhea or constipation: Some people with appendicitis experience diarrhea, while others have constipation.

Appendicitis cannot be cured via medicine alone. Antibiotics may be used to treat any accompanying infection, but they will not resolve the inflammation in the appendix. Surgery is the only effective treatment for appendicitis. The standard surgical procedure for treating appendicitis is called an appendectomy, which involves removing the appendix through an incision in the abdomen or through a small incision using laparoscopic techniques. If left untreated, appendicitis can lead to complications such as perforation of the appendix, which can result in a potentially life-threatening infection called peritonitis. Therefore, it is important to seek medical attention promptly if you suspect that you have appendicitis.

--- Question 3 ---
Q: Can you provide the trade names of medications used for treating hypertension?

Llama.generate: prefix-match hit

A: Yes, I can provide a list of some common trade names of medications used for treating hypertension (high blood pressure). Please note that this is not an exhaustive list and there are many other brand names available depending on the specific type and formulation of the medication.

1. Diuretics:
   - Hydrochlorothiazide (Microzide, Esidrix)
   - Furosemide (Lasix)
   - Spironolactone (Aldactone)
   - Chlorthalidone (Hygroton, Thalitone)

2. Beta-blockers:
   - Propranolol (Inderal)
   - Metoprolol (Lopressor, Toprol XL)
   - Atenolol (Tenormin)
   - Bisoprolol (Zebeta)

3. Calcium channel blockers:
   - Nifedipine (Procardia, Adalat CC)
   - Amlodipine (Norvasc)
   - Diltiazem (Cardizem, Tiazac)
   - Verapamil (Calan SR, Isoptin SR)

4. Angiotensin-converting enzyme (ACE) inhibitors:
   - Lisinopril (Prinivil, Zestril)
   - Enalapril (Vasotec)
   - Quinapril (Accupril)
   - Benazepril (Lotensin)

5. Angiotensin II receptor blockers (ARBs):
   - Losartan (Cozaar)
   - Valsartan (Diovan)
   - Candesartan (Atacand)
   - Olmesartan (Benicar)

6. Aldosterone antagonists:
   - Spironolactone (Aldactone)
   - Eplerenone (Inspra)

7. Renin inhibitors:
   - Aliskiren (Tasmar)

8. Vasodilators:
   - Minoxidil (Loniten)
   - Hydralazine (Apresoline)

9. Combination medications:
   - Atenolol/Chlorthalidone (Tenoretic)
   - Lisinopril/Hydrochlorothiazide (Zestoretic, Prinzide)
   - Valsartan/Sacubitril (Entresto)
   - Olmesartan/Amlodipine (Azor)

Please consult your healthcare provider for the most appropriate medication and dosage for your specific condition.

--- Question 4 ---
Q: What are the first-line options and alternatives for managing rheumatoid arthritis?

Llama.generate: prefix-match hit

A: The first-line options for managing rheumatoid arthritis (RA) typically involve a combination of medications, lifestyle modifications, and physical therapy to help reduce inflammation, alleviate symptoms, and prevent joint damage. Here are some common first-line treatments:

1. Disease-modifying antirheumatic drugs (DMARDs): These medications work by slowing down or stopping the progression of RA. Commonly used DMARDs include methotrexate, sulfasalazine, hydroxychloroquine, and leflunomide.
2. Biologic agents: For patients with severe or refractory RA, biologic agents may be considered as a first-line option. These medications target specific molecules involved in the inflammatory process and include tumor necrosis factor (TNF) inhibitors (e.g., adalimumab, etanercept), interleukin (IL)-1 inhibitors (e.g., anakinra), and B-cell targeting agents (e.g., rituximab).
3. Nonsteroidal anti-inflammatory drugs (NSAIDs): NSAIDs are often used to help manage pain and inflammation in RA. However, they do not slow down the progression of the disease like DMARDs or biologic agents. Commonly used NSAIDs include ibuprofen, naproxen, and celecoxib.
4. Corticosteroids: Corticosteroids can be used to rapidly reduce inflammation and alleviate symptoms in RA. However, they are typically not used as a first-line therapy due to their potential side effects and the risk of long-term use. Commonly used corticosteroids include prednisone and methylprednisolone.
5. Lifestyle modifications: Regular exercise, maintaining a healthy weight, and practicing stress management techniques can help improve symptoms and overall quality of life in RA patients.
6. Physical therapy: Working with a physical therapist can help improve joint mobility, strength, and flexibility, as well as reduce pain and prevent further damage to the joints.
7. Dietary modifications: Eating a healthy diet rich in fruits, vegetables, whole grains, and lean proteins may help reduce inflammation and improve overall health in RA patients. Avoiding processed foods, sugars, and refined carbohydrates can also be beneficial.
8. Complementary therapies: Acupuncture, massage therapy, and mind-body practices (e.g., meditation, yoga) may help alleviate symptoms and improve overall wellbeing in RA patients. However, these therapies should not replace conventional medical treatment.

It is important to note that the most appropriate first-line treatment for RA depends on individual factors such as disease severity, duration, and response to previous treatments. Consulting with a healthcare professional is essential to determine the best course of action for managing RA.

--- Question 5 ---
Q: What are the effective treatments or solutions for addressing sudden patchy hair loss, commonly seen as localized bald spots on the scalp, and what could be the possible causes behind it?

Llama.generate: prefix-match hit

A: Sudden patchy hair loss, also known as alopecia areata, is an autoimmune condition that causes hair loss in small patches on the scalp, beard, or other areas of the body. The exact cause of alopecia areata is not known, but it's believed to be related to an abnormal immune response that attacks the hair follicles.

There are several treatments and solutions for addressing sudden patchy hair loss:

1. Topical corticosteroids: These are anti-inflammatory medications that can help reduce inflammation in the affected area and promote hair regrowth. They can be applied directly to the bald spots in the form of creams, lotions, or solutions.
2. Injections of corticosteroids: In some cases, corticosteroids may be injected directly into the bald spots to help stimulate hair growth.
3. Immunotherapy: This involves using medications that suppress the immune system to prevent it from attacking the hair follicles. One such medication is minoxidil, which has been shown to be effective in promoting hair regrowth in some people with alopecia areata.
4. Hair transplantation: In severe cases of alopecia areata where there is extensive hair loss, hair transplantation may be an option. This involves taking healthy hair follicles from other areas of the scalp or body and transplanting them to the bald spots.
5. Lifestyle modifications: Certain lifestyle factors can contribute to hair loss, such as stress, poor nutrition, and certain medications. Making lifestyle modifications, such as reducing stress, eating a balanced diet, and avoiding medications that can cause hair loss, may help improve hair health and prevent further hair loss.

Possible causes of sudden patchy hair loss include:

1. Autoimmune conditions: Alopecia areata is an autoimmune condition where the immune system attacks the hair follicles. Other autoimmune conditions, such as lupus or rheumatoid arthritis, can also cause hair loss.
2. Stress: Physical or emotional stress can trigger hair loss in some people.
3. Nutritional deficiencies: A lack of certain nutrients, such as iron, zinc, or biotin, can contribute to hair loss.
4. Medications: Certain medications, such as chemotherapy drugs, antidepressants, and beta-blockers, can cause hair loss as a side effect.
5. Hormonal imbalances: Hormonal imbalances, such as thyroid disorders or polycystic ovary syndrome (PCOS), can cause hair loss.
6. Infections: Certain infections, such as ringworm or syphilis, can cause hair loss.
7. Trauma: Physical trauma to the scalp, such as burns or injuries, can cause hair loss.
8. Genetics: Hair loss can be inherited from both parents, and some people may be genetically predisposed to developing alopecia areata or other forms of hair loss.

--- Question 6 ---
Q: What treatments are recommended for a person who has sustained a physical injury to brain tissue, resulting in temporary or permanent impairment of brain function?

Llama.generate: prefix-match hit

# ============================================================
# Prompt Configuration Evaluation
# ============================================================

# Define 5 Configurations

configs = [
    {"name": "High Creativity", "temp": 0.9, "sys": "You are a helpful medical assistant.", "top_k": 100},
    {"name": "Strict Professional", "temp": 0.0, "sys": "You are a concise, professional doctor. Answer only what is asked.", "top_k": 10},
    {"name": "Chain of Thought", "temp": 0.3, "sys": "Think step-by-step. Explain the reasoning before giving the final answer.", "top_k": 40},
    {"name": "ELI5", "temp": 0.7, "sys": "Explain like I am 5 years old.", "top_k": 50},
    {"name": "Balanced", "temp": 0.4, "sys": "You are a knowledgeable assistant. Provide detailed clinical information.", "top_k": 50}
]

# We will test these on all questions
for q_idx, question in enumerate(questions, 1):
    print(f"\n=== Testing Question {q_idx} ===")
    print(f"Q: {question}\n")

    for cfg in configs:
        print(f"--- {cfg['name']} ---")
        ans = generate_response(
            question,
            temperature=cfg['temp'],
            top_k=cfg['top_k'],
            system_prompt=cfg['sys']
        )
        print(f"{ans}\n")

=== Testing Question 1 ===
Q: What are the common symptoms and treatments for pulmonary embolism?

--- High Creativity ---

Llama.generate: prefix-match hit

A pulmonary embolism (PE) is a serious condition caused by a blockage in one or more arteries of the lungs, usually due to a blood clot that has traveled from elsewhere in the body. The common symptoms of a PE include:

1. Shortness of breath: This is often the first symptom and can range from mild to severe. It may feel like you are unable to get enough air or that you are breathing very hard.
2. Chest pain: PE can cause sharp, stabbing chest pain, especially when taking deep breaths or coughing. The pain may be worse when lying down or bending forward.
3. Rapid heart rate: PE can cause a rapid, fluttering or irregular heartbeat.
4. Coughing up blood: In some cases, PE can cause you to cough up blood. This is a serious symptom and requires immediate medical attention.
5. Swelling: PE can cause swelling in the legs, ankles, and arms, as well as shortness of breath and chest pain.

Treatment for PE depends on the severity of the condition, but it usually involves hospitalization and medications to help dissolve blood clots or prevent new clots from forming. Some common treatments include:

1. Anticoagulants: These are medications that help prevent blood clots from forming or growing larger. They can be given as injections or taken as pills.
2. Thrombolytics: These are medications that dissolve blood clots. They are usually only used in severe cases of PE and must be administered under the close supervision of a healthcare professional.
3. Blood thinners: In some cases, a medication called a thrombin inhibitor may be used instead of an anticoagulant to help prevent further clotting.
4. Oxygen therapy: PE can cause low oxygen levels in the blood, so oxygen therapy may be needed to help keep oxygen levels stable.
5. Supportive care: PE can be a very serious condition, and supportive care such as intravenous fluids, medications to manage symptoms like pain or shortness of breath, and monitoring vital signs may be necessary.

It's important to note that if you suspect you have a PE, you should seek medical attention right away. A PE can be life-threatening, and timely treatment is essential for the best possible outcome.

--- Strict Professional ---

Llama.generate: prefix-match hit

A pulmonary embolism (PE) is a blockage in the arteries of the lungs, usually caused by a blood clot that travels from elsewhere in the body. Common symptoms include:

1. Shortness of breath
2. Chest pain or discomfort, often worsening with deep breathing or coughing
3. Rapid heart rate
4. Anxiety
5. Coughing up blood
6. Swelling in the legs or arms
7. Fainting

Treatment for PE depends on the severity of the blockage and the overall health of the patient. Options include:

1. Anticoagulant medications to prevent clot formation and growth
2. Thrombolytic therapy, which involves administering drugs that dissolve blood clots
3. Embolectomy, a surgical procedure to remove the clot from the lung artery
4. Supportive measures such as oxygen therapy, fluid replacement, and pain management.

It's important to note that PE can be life-threatening if not treated promptly, so seeking medical attention right away is crucial.

--- Chain of Thought ---

Llama.generate: prefix-match hit

A pulmonary embolism (PE) is a serious condition that occurs when a blood clot or other foreign substance blocks an artery in the lungs, obstructing blood flow and reducing oxygen supply to the body. Here are some common symptoms and treatments for PE:

Symptoms:
1. Shortness of breath: This is often the first symptom of PE and can range from mild to severe. It may feel like you're unable to get enough air, or that you're breathing too hard.
2. Chest pain: PE can cause sharp or stabbing chest pain, especially when taking deep breaths or coughing. The pain may worsen with physical activity.
3. Rapid heart rate: A fast heartbeat is a common symptom of PE and can be a sign that the body is trying to compensate for reduced oxygen supply.
4. Coughing up blood: In some cases, PE can cause coughing up blood or bloody sputum.
5. Swelling: PE can cause swelling in the legs, ankles, and feet due to increased pressure in the veins.
6. Rapid breathing: PE can cause rapid, shallow breathing as the body tries to get more oxygen.
7. Anxiety: PE can cause feelings of anxiety or fear, especially if the symptoms are severe.

Treatments:
1. Anticoagulants: These medications help prevent blood clots from forming and growing larger. They work by thinning the blood, making it harder for clots to form.
2. Thrombolytics: In some cases, PE may be treated with thrombolytic drugs, which are given intravenously and work by breaking up existing clots. This treatment is usually reserved for patients with severe symptoms or who are at high risk of complications.
3. Oxygen therapy: Patients with PE may receive oxygen therapy to help improve oxygen levels in the body.
4. Blood thinners: In some cases, PE may be treated with blood thinners, which are given intravenously and work by preventing new clots from forming.
5. Supportive care: PE can be a serious condition, and patients may require supportive care such as fluid replacement, pain relief, and monitoring of vital signs.
6. Surgery: In some cases, PE may require surgery to remove the clot or repair damage to the lungs or other organs. This is usually reserved for patients with severe symptoms or who do not respond to other treatments.
7. Lifestyle changes: Patients with PE may be advised to make lifestyle changes to reduce the risk of future clots, such as quitting smoking, getting regular exercise, and managing underlying conditions like heart disease or diabetes.

--- ELI5 ---

Llama.generate: prefix-match hit

A pulmonary embolism is when a clot, usually from your legs, travels to your lungs and blocks blood flow. This can make it hard for you to breathe and feel uncomfortable. Here are some common symptoms:

1. Trouble breathing: You might take deeper breaths than usual, or feel like you're not getting enough air.
2. Chest pain: It could feel like a sharp pain or pressure in your chest.
3. Fast heartbeat: Your heart might beat faster than normal.
4. Swelling: You might notice swelling in your legs, ankles, or feet.
5. Coughing up blood: In severe cases, you might cough up blood.

If a doctor thinks you have a pulmonary embolism, they might give you medicines to help thin your blood and make it easier for it to flow. They might also give you oxygen to help you breathe better. Sometimes people need to stay in the hospital so doctors can watch them closely and give them treatments through an IV. Other times, people can go home with medicines and follow up appointments. It's important to rest and avoid strenuous activities until you feel better.

--- Balanced ---

Llama.generate: prefix-match hit

# ============================================================
# Reference Document Ingestion
# ============================================================

# Dependencies
import requests
import os
from langchain_community.document_loaders import PyMuPDFLoader

doc_path = "014-NLP-PROJ-medical_diagnosis_manual_19.pdf"
github_url = "https://github.com/EvagAIML/014-NLP-Model-v1/blob/main/014-NLP-PROJ-medical_diagnosis_manual_19.pdf?raw=true"

# Download the PDF file from GitHub
if not os.path.exists(doc_path):
    print(f"Downloading {doc_path} from GitHub...")
    try:
        response = requests.get(github_url, stream=True)
        response.raise_for_status() # Raise an HTTPError for bad responses (4xx or 5xx)
        with open(doc_path, 'wb') as f:
            for chunk in response.iter_content(chunk_size=8192):
                f.write(chunk)
        print("Download complete.")
    except requests.exceptions.RequestException as e:
        print(f"Error downloading file from GitHub: {e}")
        print("Please ensure the URL is correct and accessible.")

if os.path.exists(doc_path):
    loader = PyMuPDFLoader(doc_path)
    docs = loader.load()
    print(f"Loaded {len(docs)} pages.")

    # Check first 5 pages
    for i in range(min(5, len(docs))):
        print(f"--- Page {i+1} ---")
        print(docs[i].page_content[:500])
        print("...")
else:
    print(f"File {doc_path} not found even after attempted download. Please ensure it is in the working directory or the download path is correct.")

Downloading 014-NLP-PROJ-medical_diagnosis_manual_19.pdf from GitHub...
Download complete.
Loaded 4114 pages.
--- Page 1 ---
erikvdesigner@gmail.com
U36PAIRLB4
ant for personal use by erikvdesigner@g
shing the contents in part or full is liable
...
--- Page 2 ---
erikvdesigner@gmail.com
U36PAIRLB4
This file is meant for personal use by erikvdesigner@gmail.com only.
Sharing or publishing the contents in part or full is liable for legal action.
...
--- Page 3 ---
Table of Contents
1
Front    ................................................................................................................................................................................................................
1
Cover    .......................................................................................................................................................................................................
2
Front Matter    .................................
...
--- Page 4 ---
491
Chapter 44. Foot & Ankle Disorders    .....................................................................................................................................
502
Chapter 45. Tumors of Bones & Joints    ...............................................................................................................................
510
5 - Ear, Nose, Throat & Dental Disorders    ........................................................................................................
...
--- Page 5 ---
921
Chapter 94. Adrenal Disorders    ................................................................................................................................................
936
Chapter 95. Polyglandular Deficiency Syndromes    ........................................................................................................
939
Chapter 96. Porphyrias    .................................................................................................................................
...

text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=1000,
    chunk_overlap=200
)
chunks = text_splitter.split_documents(docs)
print(f"Total Chunks created: {len(chunks)}")

Total Chunks created: 18020

# ============================================================
# Embedding & Vector Store
# ============================================================

from langchain_community.embeddings.sentence_transformer import SentenceTransformerEmbeddings

embedding_model = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")

vector_db = Chroma.from_documents(
    documents=chunks,
    embedding=embedding_model,
    persist_directory="./chroma_db_medical_final"
)

# Default Retriever
retriever = vector_db.as_retriever(search_kwargs={"k": 3})

/tmp/ipython-input-4154461508.py:7: LangChainDeprecationWarning: The class `HuggingFaceEmbeddings` was deprecated in LangChain 0.2.2 and will be removed in 1.0. An updated version of the class exists in the :class:`~langchain-huggingface package and should be used instead. To use it run `pip install -U :class:`~langchain-huggingface` and import as `from :class:`~langchain_huggingface import HuggingFaceEmbeddings``.
  embedding_model = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")

modules.json:   0%|          | 0.00/349 [00:00<?, ?B/s]

config_sentence_transformers.json:   0%|          | 0.00/116 [00:00<?, ?B/s]

README.md: 0.00B [00:00, ?B/s]

sentence_bert_config.json:   0%|          | 0.00/53.0 [00:00<?, ?B/s]

config.json:   0%|          | 0.00/612 [00:00<?, ?B/s]

model.safetensors:   0%|          | 0.00/90.9M [00:00<?, ?B/s]

tokenizer_config.json:   0%|          | 0.00/350 [00:00<?, ?B/s]

vocab.txt: 0.00B [00:00, ?B/s]

tokenizer.json: 0.00B [00:00, ?B/s]

# ============================================================
# RAG Generation Function
# ============================================================

def generate_rag_answer(query, k=3, temperature=0.1, max_tokens=2048):
    # Retrieve
    retrieved_docs = vector_db.similarity_search(query, k=k)

    # Format context and extract sources
    context_list = []
    sources_info = []
    for d in retrieved_docs:
        page_num = d.metadata.get('page', 'Unknown')
        # Assuming page is 0-indexed integer from PyMuPDF
        if isinstance(page_num, int):
            page_num += 1
        src_str = f"Page {page_num}"
        sources_info.append(src_str)
        context_list.append(f"[{src_str}] {d.page_content}")

    context = "\n\n".join(context_list)
    unique_sources = ", ".join(sorted(list(set(sources_info))))

    # Construct Prompt
    sys_msg = "You are a professional medical assistant for healthcare professionals. Use the following Context to answer the Question. If the answer is not in the context, say so. Do not advise consulting a healthcare professional."
    user_msg = f"Context:\n{context}\n\nQuestion: {query}\n\nPlease cite specific Page numbers and Sections from the text in your answer."

    prompt = f"[INST] {sys_msg}\n{user_msg} [/INST]"

    # Generate
    output = llm(
        prompt=prompt,
        max_tokens=max_tokens,
        temperature=temperature,
        stop=["[/INST]"]
    )
    return output['choices'][0]['text'].strip(), context

# ============================================================
# RAG Parameter Tuning (5 Combinations)
# ============================================================

rag_configs = [
    {"name": "C1: Low Context, Precise", "k": 2, "temp": 0.0},
    {"name": "C2: Standard", "k": 3, "temp": 0.1},
    {"name": "C3: High Context", "k": 5, "temp": 0.1},
    {"name": "C4: Creative", "k": 3, "temp": 0.7},
    {"name": "C5: Max Context, Strict", "k": 7, "temp": 0.0}
]

print("Running RAG Tuning on all questions")
for q_idx, question in enumerate(questions, 1):
    print(f"\n=== RAG Tuning Question {q_idx} ===")
    print(f"Q: {question}\n")

    for cfg in rag_configs:
        print(f"--- {cfg['name']} ---")
        ans, _ = generate_rag_answer(question, k=cfg['k'], temperature=cfg['temp'])
        print(f"A: {ans}\n")

Running RAG Tuning on all questions

=== RAG Tuning Question 1 ===
Q: What are the common symptoms and treatments for pulmonary embolism?

--- C1: Low Context, Precise ---

Llama.generate: prefix-match hit

A: According to the context on pages 2080 and 2081 of Chapter 194, the common symptoms for pulmonary embolism include dyspnea (shortness of breath), pleuritic chest pain, cough, syncope or cardiorespiratory arrest in severe cases, tachypnea (rapid breathing), tachycardia (fast heart rate), hypotension (low blood pressure), and a loud pulmonic component of the 2nd heart sound.

The treatment for pulmonary embolism includes anticoagulants to prevent clot formation and growth, and in some cases, clot dissolution with thrombolytics or surgical removal. Preventive measures include anticoagulants and the insertion of an inferior vena caval filter (Section: "Treatment" on page 2080).

Therefore, the answer is: The common symptoms for pulmonary embolism are dyspnea, pleuritic chest pain, cough, syncope or cardiorespiratory arrest in severe cases, tachypnea, tachycardia, hypotension, and a loud pulmonic component of the 2nd heart sound (Page 2080). The treatment includes anticoagulants, thrombolytics, surgical removal, and preventive measures such as anticoagulants and inferior vena caval filters (Page 2080).

--- C2: Standard ---

Llama.generate: prefix-match hit

A: According to the context on pages 2080 and 2081, the common symptoms of pulmonary embolism include dyspnea, pleuritic chest pain, cough, tachypnea, tachycardia, hypotension, and a loud pulmonic component of the 2nd heart sound. (Section: Introduction)

The treatment for pulmonary embolism involves anticoagulants and, in some cases, clot dissolution with thrombolytics or surgical removal. Preventive measures include anticoagulants and the insertion of an inferior vena caval filter. (Section: Treatment)

Please note that this information is based on the provided context and may not cover all possible symptoms or treatments for pulmonary embolism. For comprehensive medical advice, consult a healthcare professional. (I cannot provide such advice.)

--- C3: High Context ---

Llama.generate: prefix-match hit

A: The common symptoms for pulmonary embolism (PE) include dyspnea, pleuritic chest pain, cough, tachypnea, tachycardia, hypotension, and a loud pulmonic component of the 2nd heart sound (Page 2080). The diagnosis is based on imaging tests such as CT angiogram, ventilation/perfusion scan, or pulmonary arteriogram (Page 2080).

The treatment for PE includes anticoagulants and, in some cases, clot dissolution with thrombolytics or surgical removal (Page 2080). Preventive measures include anticoagulants and the insertion of an inferior vena caval filter (Page 2080).

It is important to note that hemoptysis, which is a symptom mentioned in the text, is not a common initial manifestation of PE. Instead, it is more commonly caused by bronchitis, bronchiectasis, TB, and necrotizing pneumonia or lung abscess (Page 2017).

Therefore, the symptoms and treatments specifically for pulmonary embolism are mentioned on Page 2080 of the text.

--- C4: Creative ---

Llama.generate: prefix-match hit

A: According to the context on pages 2080 and 2081, the common symptoms for pulmonary embolism include dyspnea, pleuritic chest pain, cough, tachypnea, tachycardia, hypotension, syncope or cardiorespiratory arrest, and a loud pulmonic component of the 2nd heart sound. (Section: Introduction)

The treatment for pulmonary embolism includes anticoagulants, clot dissolution with thrombolytics, surgical removal, and preventive measures like anticoagulants and inferior vena caval filters. (Section: Treatment)

Please note that this answer is based on the information provided in the context, and it's essential to consult a healthcare professional for accurate diagnosis and treatment. (Disclaimer: I'm just a helpful AI, not a doctor!)

--- C5: Max Context, Strict ---

Llama.generate: prefix-match hit

# ============================================================
# Final RAG Configuration -> Answers
# ============================================================

rag_responses = []
rag_contexts = []

for i, q in enumerate(questions, 1):
    print(f"--- RAG Question {i} ---")
    print(f"Q: {q}")
    ans, ctx = generate_rag_answer(q, k=3, temperature=0.1)
    print(f"A: {ans}\n")
    rag_responses.append(ans)
    rag_contexts.append(ctx)

--- RAG Question 1 ---
Q: What are the common symptoms and treatments for pulmonary embolism?

Llama.generate: prefix-match hit

A: According to the context on pages 2080 and 2081, the common symptoms of pulmonary embolism include dyspnea, pleuritic chest pain, cough, tachypnea, tachycardia, hypotension, and a loud pulmonic component of the 2nd heart sound. (Section: Introduction)

The treatment for pulmonary embolism involves anticoagulants and, in some cases, clot dissolution with thrombolytics or surgical removal. Preventive measures include anticoagulants and inferior vena caval filters. (Section: Treatment)

Please note that the context also mentions other conditions such as pneumonia, sepsis, foreign body embolism, tumor embolism, pulmonary-renal syndromes, and chronic illnesses that may present with similar symptoms or signs. It is essential to consider these possibilities in the diagnostic workup. (Sections: signs of pneumonia or sepsis, Foreign body embolism, Tumor embolism, Several specific findings are of note)

--- RAG Question 2 ---
Q: What are the common symptoms for appendicitis, and can it be cured via medicine? If not, what surgical procedure should be followed to treat it?

Llama.generate: prefix-match hit

A: The common symptoms of appendicitis include epigastric or periumbilical pain followed by brief nausea, vomiting, and anorexia, which is then followed by a shift in pain to the right lower quadrant. Pain increases with cough and motion. Classic signs are direct and rebound tenderness located at McBurney's point (junction of the middle and outer thirds of the line joining the umbilicus to the anterior superior spine). Additional signs are pain felt in the right lower quadrant with palpation of the left lower quadrant (Merck Manual, p. 174, Symptoms and Signs section).

Appendicitis cannot be cured via medicine alone as it requires surgical removal to prevent complications such as necrosis, gangrene, perforation, abscess formation, or inflammatory masses (Merck Manual, p. 174, Etiology section). The surgical procedure for treating appendicitis is typically an appendectomy, which may involve removing the appendix alone if it's not yet ruptured or draining an abscess if one has formed (Merck Manual, p. 174, Treatment section).

Before the operation, an NGT (nasogastric tube) is inserted, and patients with signs of volume depletion should have urine output monitored with a catheter. Fluid status is maintained by adequate IV fluid and electrolyte replacement, and IV antibiotics effective against intestinal flora should be given (Merck Manual, p. 174, Treatment section).

--- RAG Question 3 ---
Q: Can you provide the trade names of medications used for treating hypertension?

Llama.generate: prefix-match hit

A: The context does mention some specific medications used for treating hypertension, but it does not provide their trade names. The medications mentioned are:

1. ACE inhibitors (Page 2237)
2. Angiotensin II receptor blockers (Page 2237)
3. Thiazide-type diuretics, loop diuretics, and K-sparing diuretics (Page 2236 and 2238)
4. Prazosin, terazosin, doxazosin (Page 2238)
5. Reserpine, guanethidine, guanadrel (Page 2238)

To find the trade names of these medications, I would recommend checking a reliable drug database or consulting a healthcare professional. The Merck Manual does not provide the trade names in the given context.

--- RAG Question 4 ---
Q: What are the first-line options and alternatives for managing rheumatoid arthritis?

Llama.generate: prefix-match hit

A: According to the context on pages 426 and 430 of "Geriatrics Essentials" and page 382 of "Chapter 35. Joint Disorders," the first-line options for managing rheumatoid arthritis (RA) include:

1. Disease-modifying antirheumatic drugs (DMARDs): These medications help control symptoms and slow disease progression. They are used early, often in combination (Page 426).

Alternatives for managing RA, as mentioned in the context, include:

1. Symptomatic treatment for joint inflammation: Nonsteroidal anti-inflammatory drugs (NSAIDs) are typically used to treat joint inflammation (Page 382).
2. Pain management without inflammation: Acetaminophen is a safer option for pain relief in the absence of inflammation (Page 382).
3. Joint immobilization: Splints or slings can help relieve pain by immobilizing affected joints (Page 382).
4. Heat and cold therapy: Heat therapy may help relieve muscle spasms around joints, while cold therapy can be analgesic in inflammatory joint diseases (Page 382).
5. Surgery: Arthroplasty with prosthetic joint replacement is indicated when damage severely limits function, such as in total hip and knee replacements (Page 430). Other surgical options include excision of subluxed painful metatarsophalangeal joints or thumb fusions to aid walking and provide stability for pinch, respectively. Neck fusion may be necessary for C1-2 subluxation with severe pain or potential for spinal cord compression (Page 430). Arthroscopic or open synovectomy can temporarily relieve joint inflammation but only if disease activity can be controlled (Page 430).

It is essential to note that the context does not provide specific information on alternative medications for managing RA beyond DMARDs and NSAIDs. Always consult a healthcare professional for personalized medical advice.

--- RAG Question 5 ---
Q: What are the effective treatments or solutions for addressing sudden patchy hair loss, commonly seen as localized bald spots on the scalp, and what could be the possible causes behind it?

Llama.generate: prefix-match hit

# ============================================================
# Response Evaluation (Refined 1-5 Scoring with evidence)
# ============================================================

import re

groundedness_rater_system_message = """
You are tasked with rating AI generated answers to questions posed by users.
You will be presented a question, context used by the AI system to generate the answer and an AI generated answer to the question.
In the input, the question will begin with ###Question, the context will begin with ###Context while the AI generated answer will begin with ###Answer.

Evaluation criteria:
The task is to judge the extent to which the metric is followed by the answer.
1 - The metric is not followed at all
2 - The metric is followed only to a limited extent
3 - The metric is followed to a good extent
4 - The metric is followed mostly
5 - The metric is followed completely

Metric:
The answer should be derived only from the information presented in the context

Instructions:
1. First write down the steps that are needed to evaluate the answer as per the metric.
2. Give a step-by-step explanation if the answer adheres to the metric considering the question and context as the input.
3. Next, evaluate the extent to which the metric is followed.
4. Use the previous information to rate the answer using the evaluaton criteria and assign a score.
5. IMPORTANT: End your response with a single line: SCORE: X (where X is 1, 2, 3, 4, or 5)
"""

relevance_rater_system_message = """
You are tasked with rating AI generated answers to questions posed by users.
You will be presented a question, context used by the AI system to generate the answer and an AI generated answer to the question.
In the input, the question will begin with ###Question, the context will begin with ###Context while the AI generated answer will begin with ###Answer.

Evaluation criteria:
The task is to judge the extent to which the metric is followed by the answer.
1 - The metric is not followed at all
2 - The metric is followed only to a limited extent
3 - The metric is followed to a good extent
4 - The metric is followed mostly
5 - The metric is followed completely

Metric:
Relevance measures how well the answer addresses the main aspects of the question, based on the context.
Consider whether all and only the important aspects are contained in the answer when evaluating relevance.

Instructions:
1. First write down the steps that are needed to evaluate the context as per the metric.
2. Give a step-by-step explanation if the context adheres to the metric considering the question as the input.
3. Next, evaluate the extent to which the metric is followed.
4. Use the previous information to rate the context using the evaluaton criteria and assign a score.
5. IMPORTANT: End your response with a single line: SCORE: X (where X is 1, 2, 3, 4, or 5)
"""

def evaluate_response(query, answer, context):
    # Groundedness Prompt
    ground_prompt = f"[INST] {groundedness_rater_system_message}\n\n###Question\n{query}\n###Context\n{context}\n###Answer\n{answer}\n[/INST]"

    # Use Main LLM (Mistral)
    ground_out = llm(ground_prompt, max_tokens=512, temperature=0.1)['choices'][0]['text'].strip()

    # Relevance Prompt
    rel_prompt = f"[INST] {relevance_rater_system_message}\n\n###Question\n{query}\n###Context\n{context}\n###Answer\n{answer}\n[/INST]"

    # Use Main LLM (Mistral)
    rel_out = llm(rel_prompt, max_tokens=512, temperature=0.1)['choices'][0]['text'].strip()

    # Parse Scores using Regex (Prioritize explicit format)
    def parse_score(text):
        # Try strict format first: SCORE: 5
        strict_matches = re.findall(r'SCORE:\s*([1-5])', text, re.IGNORECASE)
        if strict_matches:
            return int(strict_matches[-1])

        # Fallback: Look for last digit 1-5
        matches = re.findall(r'\b([1-5])\b', text)
        if matches:
            return int(matches[-1])
        return 0 # Error or not found

    ground_score = parse_score(ground_out)
    rel_score = parse_score(rel_out)

    return ground_score, ground_out, rel_score, rel_out

print("--- Evaluation Results ---\n")
for i, (q, a, c) in enumerate(zip(questions, rag_responses, rag_contexts), 1):
    gs, ge, rs, re_text = evaluate_response(q, a, c)
    print(f"Q{i}: {q}")
    print(f"Groundedness Score: {gs}/5")
    print(f"Reason: {ge}")
    print(f"Relevance Score: {rs}/5")
    print(f"Reason: {re_text}")
    print("-" * 50)

--- Evaluation Results ---

Llama.generate: prefix-match hit
Llama.generate: prefix-match hit

Q1: What are the common symptoms and treatments for pulmonary embolism?
Groundedness Score: 4/5
Reason: Steps for evaluation:
1. Identify the common symptoms and treatments mentioned in the context related to pulmonary embolism.
2. Compare the identified information with the AI generated answer.
3. Determine if the AI generated answer is derived only from the context without any additional information.

Explanation:
The AI generated answer adheres to the metric as it mentions the common symptoms and treatments for pulmonary embolism based on the context provided. The answer does not include any additional information.

Evaluation:
The metric is followed mostly (4) since the answer includes all the required information from the context but may lack a few minor details.

SCORE: 4
Relevance Score: 4/5
Reason: Steps to evaluate the context as per the metric:
1. Read the entire context carefully to understand the main aspects of pulmonary embolism.
2. Identify the sections that discuss common symptoms and treatments for pulmonary embolism.
3. Extract the relevant information about common symptoms and treatments from these sections.
4. Consider how well the extracted information addresses the main aspects of the question.
5. Evaluate to what extent the context adheres to the metric by assessing if all and only the important aspects are contained in the answer.

The context adheres to the metric considering the question as the input because it provides detailed information about common symptoms (dyspnea, pleuritic chest pain, cough, tachypnea, tachycardia, hypotension, and a loud pulmonic component of the 2nd heart sound) and treatments (anticoagulants, clot dissolution with thrombolytics, surgical removal, and preventive measures such as anticoagulants and inferior vena caval filters) for pulmonary embolism. The context also acknowledges other conditions that may present with similar symptoms or signs but emphasizes the importance of considering these possibilities in the diagnostic workup.

The metric is followed mostly because the context provides a comprehensive overview of common symptoms, treatments, and potential confounding conditions related to pulmonary embolism. It covers all essential aspects of the question while also acknowledging the importance of considering other possible diagnoses.

SCORE: 4
--------------------------------------------------

Llama.generate: prefix-match hit
Llama.generate: prefix-match hit

Q2: What are the common symptoms for appendicitis, and can it be cured via medicine? If not, what surgical procedure should be followed to treat it?
Groundedness Score: 5/5
Reason: Steps to evaluate the answer:
1. Identify the common symptoms of appendicitis mentioned in the context.
2. Check if the AI generated answer includes those symptoms.
3. Determine if the AI generated answer mentions that appendicitis cannot be cured via medicine and requires surgical removal.
4. Verify if the surgical procedure for treating appendicitis, as stated in the answer, is consistent with what's mentioned in the context.
5. Ensure that all information provided in the answer can be traced back to the context.

The AI generated answer adheres to the metric since it only includes information derived from the context. The symptoms and surgical procedure mentioned in the answer are directly taken from the context, and the statement about not being able to cure appendicitis with medicine is also consistent with the information provided in the context.

Evaluation: The metric is followed completely.

SCORE: 5
Relevance Score: 4/5
Reason: Steps to evaluate the context as per the metric:
1. Identify the main aspects of the question: common symptoms for appendicitis and whether it can be cured via medicine or if surgery is required.
2. Read through the context to find information related to the question's main aspects.
3. Determine how well the context addresses each main aspect of the question.
4. Evaluate the relevance of the context based on the extent to which it covers all and only the important aspects of the question.

The context adheres to the metric considering the question as the input:
1. The context provides detailed information about the symptoms and signs of appendicitis, making it highly relevant to the first main aspect of the question.
2. The context also mentions that appendicitis cannot be cured via medicine alone and requires surgical removal, which is directly related to the second main aspect of the question.
3. Furthermore, the context explains the different procedures for treating appendicitis, such as an appendectomy or draining an abscess, making it relevant to both parts of the question.

The metric is followed mostly:
1. The context provides comprehensive information about the symptoms and signs of appendicitis, which are directly related to the first main aspect of the question.
2. It also covers the fact that appendicitis cannot be cured via medicine alone and requires surgical removal, addressing the second main aspect of the question.
3. The context further explains the different procedures for treating appendicitis, providing additional relevant information.
4. However, it does not mention specific medications that could potentially cure appendicitis, which is a limitation in fully following the metric.

SCORE: 4
--------------------------------------------------

Llama.generate: prefix-match hit
Llama.generate: prefix-match hit

Q3: Can you provide the trade names of medications used for treating hypertension?
Groundedness Score: 5/5
Reason: Steps to evaluate the answer:
1. Identify the specific medications mentioned in the context.
2. Check if the context provides the trade names of these medications.
3. If the context does not provide the trade names, suggest checking a reliable drug database or consulting a healthcare professional for the information.

The answer adheres to the metric as it is derived only from the information presented in the context and does not provide any additional information or speculation. The answer suggests looking up the trade names elsewhere due to the lack of this information in the context.

Extent to which the metric is followed: 5 (The answer is completely following the metric by deriving the information only from the context)

SCORE: 5
Relevance Score: 2/5
Reason: Steps to evaluate the context as per the metric:
1. Identify the main aspects of the question which are the trade names of medications used for treating hypertension.
2. Determine if the context contains all and only the important aspects related to the question.
3. Check if the context provides the trade names of the mentioned medications.

The context adheres to the metric considering the question as the input to a limited extent because it mentions some specific medications used for treating hypertension but does not provide their trade names. SCORE: 2
--------------------------------------------------

Llama.generate: prefix-match hit
Llama.generate: prefix-match hit

Q4: What are the first-line options and alternatives for managing rheumatoid arthritis?
Groundedness Score: 5/5
Reason: STEPS TO EVALUATE ANSWER:
1. Read the question carefully.
2. Identify the key terms in the question.
3. Locate the context related to the question.
4. Extract relevant information from the context.
5. Use the extracted information to generate an answer that directly addresses the question.
6. Ensure the answer is derived only from the information presented in the context.

ANALYSIS:
The answer adheres to the metric as it is generated solely based on the information provided in the context. The answer lists the first-line options and alternatives for managing rheumatoid arthritis (RA) directly from the context, without introducing any external or irrelevant information.

RATING:
5 - The metric is followed completely.

SCORE: 5
Relevance Score: 4/5
Reason: Steps to evaluate the context as per the metric:
1. Identify the main aspects of the question, which are the first-line options and alternatives for managing rheumatoid arthritis.
2. Review the context to determine if it addresses all and only the important aspects related to the question.
3. Check if the context provides sufficient information about the first-line options and alternatives mentioned in the question.

The context adheres to the metric considering the question as the input because:
1. The context discusses both the first-line options (disease-modifying antirheumatic drugs, or DMARDs, and symptomatic treatment) and alternatives (NSAIDs for pain relief without inflammation, joint immobilization with splints or slings, heat therapy, cold therapy, surgery, and arthroscopic or open synovectomy) for managing rheumatoid arthritis.
2. The context provides detailed information about each option or alternative, explaining their roles and benefits in managing RA.
3. The context also mentions the importance of consulting a healthcare professional for personalized medical advice.

The metric is followed mostly because the context covers all the important aspects related to the question while providing sufficient detail about each first-line option and alternative for managing rheumatoid arthritis.

SCORE: 4
--------------------------------------------------

Llama.generate: prefix-match hit
Llama.generate: prefix-match hit

Medical Diagnostic Assistant: AI-Powered Question Answering¶

Executive Summary¶

Problem Statement¶

Business Context¶

Objective¶

Data Description¶

Environment Initialization¶

Environment Setup and Dependency Configuration¶

Core Imports and Shared Configuration¶

Baseline Medical Question Answering (Ungrounded)¶

Baseline Model Initialization¶

Inference Wrapper¶

Baseline Question Execution¶

Baseline Analysis¶

Prompt Engineering and Decoding Control (Parameter Tuning)¶

Prompt Configuration Evaluation¶

Prompting Analysis¶

Reference Preparation for Retrieval-Augmented Generation¶

Reference Document Ingestion¶

Reference Text Chunking¶

Vector Index Construction¶

Retrieval Preparation Analysis¶

Retrieval-Augmented Medical Question Answering¶

RAG Answer Generation¶

Grounded Question Execution¶

Final RAG Answers (Best Configuration: k=3, temp=0.1)¶

RAG Analysis¶

Automated Output Evaluation (LLM as the judge)¶

Response Scoring¶

Summary: Business value and deployment strategy defined¶

Next Steps: Model and Data Enhancements¶

Expanded Executive Summary¶