push to huggingface space

.env

@@ -0,0 +1 @@
+HF_TOKEN=

README.md

@@ -1,13 +1,33 @@
----
-title: MedicalSearchModel
-emoji: 🌍
-colorFrom: gray
-colorTo: blue
-sdk: gradio
-sdk_version: 6.5.1
-app_file: app.py
-pinned: false
-short_description: Interactive tool to test "embeddinggemma-300m-medical"
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# 🔍 Medical Document Search using sentence-transformers/embeddinggemma-300m-medical
+
+A search tool specialized in the medical field to helps you find relevant information across your medical documents.
+
+## How It Works
+
+1. **Enter your question** in the reference sentence box (e.g., "Is Mr. Allen eligible for enrollment given his type 2 diabetes?")
+2. **Add documents** to search through in the comparison sentence boxes
+3. **Click "Calculate Similarity"** to see ranked results
+4. **Review the scores**: 
+   - 🟢 Green (≥0.70): High similarity - very relevant
+   - 🟠 Orange (0.50-0.69): Medium similarity - somewhat relevant
+   - 🟣 Purple (<0.50): Lower similarity - less relevant
+
+## Installation
+```bash
+pip install -r requirements.txt
+```
+
+## Setup
+
+1. Get a Hugging Face API token from [huggingface.co/settings/tokens](https://huggingface.co/settings/tokens)
+2. Create a `.env` file in the project directory:
+```
+   HF_TOKEN=your_token_here
+```
+3. Run the application:
+```bash
+   python main.py
+```
+
+## Notes 
+Note that for efficiency purposes, a template of the interface code and the css code was generate with an LLM. 

config.py

@@ -0,0 +1,6 @@
+import os
+from dotenv import load_dotenv
+
+load_dotenv()
+
+HF_TOKEN = os.environ.get("HF_TOKEN")

custom_css.py

@@ -0,0 +1,125 @@
+CUSTOM_CSS_FOR_INTERFACE = """
+#title {
+    text-align: center;
+    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+    color: white;
+    padding: 40px 20px;
+    border-radius: 16px;
+    margin-bottom: 30px;
+}
+#title h1 {
+    margin: 0;
+    font-size: 2.5em;
+    font-weight: 700;
+}
+#subtitle {
+    margin-top: 10px;
+    opacity: 0.95;
+    font-size: 1.1em;
+}
+.input-section {
+    background: #f7fafc;
+    padding: 24px;
+    border-radius: 12px;
+    border: 1px solid #e2e8f0;
+}
+#calc-btn {
+    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%) !important;
+    border: none !important;
+    font-size: 16px !important;
+    font-weight: 600 !important;
+    padding: 12px 32px !important;
+    border-radius: 8px !important;
+    transition: transform 0.2s ease !important;
+}
+#calc-btn:hover {
+    transform: translateY(-2px);
+    box-shadow: 0 8px 16px rgba(102, 126, 234, 0.3) !important;
+}
+.add-btn {
+    background: #48bb78 !important;
+    border: none !important;
+    color: white !important;
+    font-weight: 600 !important;
+    border-radius: 8px !important;
+    margin-top: 12px !important;
+}
+.add-btn:hover {
+    background: #38a169 !important;
+}
+.comparison-box {
+    margin-bottom: 8px;
+}
+"""
+
+SIMILARITY_BARS = """
+<style>
+    .similarity-container {
+        font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif;
+    }
+    .similarity-item {
+        margin-bottom: 24px;
+        padding: 16px;
+        background: linear-gradient(to right, #f8f9fa 0%, #ffffff 100%);
+        border-radius: 12px;
+        border-left: 4px solid #667eea;
+        transition: transform 0.2s ease;
+    }
+    .similarity-item:hover {
+        transform: translateX(4px);
+        box-shadow: 0 4px 12px rgba(102, 126, 234, 0.15);
+    }
+    .sentence-text {
+        font-size: 15px;
+        color: #2d3748;
+        margin-bottom: 12px;
+        line-height: 1.6;
+        font-weight: 500;
+    }
+    .bar-container {
+        display: flex;
+        align-items: center;
+        gap: 12px;
+    }
+    .progress-bar {
+        flex: 1;
+        height: 28px;
+        background: #e2e8f0;
+        border-radius: 14px;
+        overflow: hidden;
+        position: relative;
+    }
+    .progress-fill {
+        height: 100%;
+        background: linear-gradient(90deg, #667eea 0%, #764ba2 100%);
+        border-radius: 14px;
+        transition: width 0.6s ease;
+        display: flex;
+        align-items: center;
+        justify-content: flex-end;
+        padding-right: 10px;
+    }
+    .score-badge {
+        background: #667eea;
+        color: white;
+        padding: 6px 16px;
+        border-radius: 20px;
+        font-weight: 600;
+        font-size: 14px;
+        min-width: 70px;
+        text-align: center;
+        box-shadow: 0 2px 8px rgba(102, 126, 234, 0.3);
+    }
+    .rank-badge {
+        display: inline-block;
+        background: #f7fafc;
+        color: #4a5568;
+        padding: 4px 10px;
+        border-radius: 8px;
+        font-size: 12px;
+        font-weight: 600;
+        margin-right: 8px;
+    }
+</style>
+<div class="similarity-container">
+"""

default_values.py

@@ -0,0 +1,9 @@
+DEFAULT_DOCUMENTS = [
+                        "Patients with diabetes are not eligible for the study.",
+                        "Participants with moderate hepatic impairment received Drug 1 at a reduced dose of 100 mg once daily.",
+                        "Patients with a history of alcoholic liver disease within the past 5 years were excluded from participation in the study.",
+                        "Inclusion criteria required a liver biopsy performed within 6 months.",
+                        "Drug 1 was administered orally once daily for a total treatment duration of 48 weeks.",
+                    ]
+
+DEFAULT_QUESTION = "Is Mr. Allen eligible for enrollment given his type 2 diabetes?"

interface.py

@@ -0,0 +1,76 @@
+import gradio as gr
+from similarity import compute_similarity
+from custom_css import CUSTOM_CSS_FOR_INTERFACE
+from default_values import DEFAULT_DOCUMENTS, DEFAULT_QUESTION
+
+
+def build_interface():
+    """Build and return the Gradio interface"""
+    with gr.Blocks(css=CUSTOM_CSS_FOR_INTERFACE, theme=gr.themes.Soft()) as demo:
+        with gr.Column(elem_id="title"):
+            gr.HTML("<h1>🔍 Smart Medical Search</h1>")
+            gr.HTML("<h2> Ask questions, get instant answers from your documents</h2>")
+
+        with gr.Row():
+            with gr.Column(scale=1):
+                with gr.Group(elem_classes="input-section"):
+                    reference_input = gr.Textbox(
+                        label="📌 Enter your question about your medical documents:",
+                        placeholder="Enter your question here...",
+                        value=DEFAULT_QUESTION,
+                        lines=3,
+                        max_lines=5
+                    )
+
+                    gr.HTML("""<h3> 🔍  Your medical documents:</h3>""")
+
+                    comparison_container = gr.Column()
+
+                    with comparison_container:
+                        comparison_inputs = []
+                        for i in range(15):
+                            value = DEFAULT_DOCUMENTS[i] if i < len(DEFAULT_DOCUMENTS) else ""
+                            comparison_box = gr.Textbox(
+                                label=f"Sentence {i + 1}",
+                                placeholder="Enter a document to search...",
+                                value=value,
+                                lines=2,
+                                elem_classes="comparison-box",
+                                visible=True if i < 5 else False
+                            )
+                            comparison_inputs.append(comparison_box)
+
+
+                    add_btn = gr.Button("➕ Add another document", elem_classes="add-btn", size="md")
+
+                    visible_count = gr.State(5)
+
+                    add_btn.click(
+                        fn=_add_comparison_box,
+                        inputs=[visible_count],
+                        outputs=[visible_count] + comparison_inputs
+                    )
+
+                submit_btn = gr.Button("Search for the answer in your documents", variant="primary", elem_id="calc-btn", size="lg")
+
+            with gr.Column(scale=1):
+                output = gr.HTML(label="Similarity Scores")
+
+        submit_btn.click(
+            fn=compute_similarity,
+            inputs=[reference_input] + comparison_inputs,
+            outputs=output
+        )
+
+    return demo
+
+
+def _add_comparison_box(count: int) -> list[int]:
+    """Check whether new boxes can be added."""
+    if count < 15:
+        new_count = count + 1
+        visibility = [gr.update(visible=True) if i < new_count else gr.update(visible=False)
+                      for i in range(15)]
+        return [new_count] + visibility
+    else:
+        return [count] + [gr.update() for _ in range(15)]

main.py

@@ -0,0 +1,5 @@
+from interface import build_interface
+
+if __name__ == "__main__":
+    demo = build_interface()
+    demo.launch()

requirements.txt

@@ -0,0 +1,3 @@
+gradio
+huggingface_hub
+python-dotenv

similarity.py

@@ -0,0 +1,57 @@
+from huggingface_hub import InferenceClient
+from config import HF_TOKEN
+from custom_css import SIMILARITY_BARS
+
+client = InferenceClient(
+    provider="hf-inference",
+    api_key=HF_TOKEN,
+)
+
+
+
+def compute_similarity(reference_sentence: str, *comparison_sentences):
+    """Compute similarity scores between reference and comparison sentences."""
+    sentences_list = [s.strip() for s in comparison_sentences if s and s.strip()]
+
+    if not sentences_list:
+        missing_reference_documents_message = "⚠️ Please enter at least one comparison sentence."
+        return f"<div style='color: #e74c3c; padding: 20px; text-align: center;'>{missing_reference_documents_message}</div>"
+
+    try:
+        result = client.sentence_similarity(
+            sentence=reference_sentence,
+            other_sentences=sentences_list,
+            model="sentence-transformers/embeddinggemma-300m-medical",
+        )
+        sorted_results = sorted(zip(sentences_list, result), key=lambda x: x[1], reverse=True)
+        output = SIMILARITY_BARS
+
+        for idx, (sentence, score) in enumerate(sorted_results, 1):
+            percentage = score * 100
+            if score >= 0.7:
+                border_color = "#48bb78"
+                gradient = "linear-gradient(90deg, #48bb78 0%, #38a169 100%)"
+            elif score >= 0.5:
+                border_color = "#ed8936"
+                gradient = "linear-gradient(90deg, #ed8936 0%, #dd6b20 100%)"
+            else:
+                border_color = "#667eea"
+                gradient = "linear-gradient(90deg, #667eea 0%, #764ba2 100%)"
+
+            output += f"""
+            <div class="similarity-item" style="border-left-color: {border_color};">
+                <div class="sentence-text">
+                    <span class="rank-badge">#{idx}</span>{sentence}
+                </div>
+                <div class="bar-container">
+                    <div class="progress-bar">
+                        <div class="progress-fill" style="width: {percentage}%; background: {gradient};"></div>
+                    </div>
+                    <div class="score-badge" style="background: {border_color};">{score:.4f}</div>
+                </div>
+            </div>
+            """
+        output += "</div>"
+        return output
+    except Exception as e:
+        return f"<div style='color: #e74c3c; padding: 20px; background: #fee; border-radius: 8px; border-left: 4px solid #e74c3c;'><strong>Error:</strong> {str(e)}</div>"