app.py

import os
os.environ['ANONYMIZED_TELEMETRY'] = 'False'

import zipfile
import chromadb
from sentence_transformers import SentenceTransformer
import gradio as gr
from fastapi import FastAPI
from pydantic import BaseModel

# Extract and load database
DB_PATH = "./medqa_db"
if not os.path.exists(DB_PATH) and os.path.exists("./medqa_db.zip"):
    with zipfile.ZipFile("./medqa_db.zip", 'r') as z:
        z.extractall(".")

client = chromadb.PersistentClient(path=DB_PATH)
collection = client.get_collection("medqa")
model = SentenceTransformer('ncbi/MedCPT-Query-Encoder')

# Search function
def search(query, num_results=3):
    emb = model.encode(query).tolist()
    return collection.query(query_embeddings=[emb], n_results=int(num_results))

# Gradio UI
def ui_search(query, num_results=3):
    if not query.strip():
        return "Enter a query"
    try:
        r = search(query, num_results)
        out = ""
        for i in range(len(r['documents'][0])):
            out += f"\n{'='*60}\nExample {i+1}\n{'='*60}\n"
            
            # Get the full question text
            question_text = r['documents'][0][i]
            
            # DEBUG: Show raw text
            out += "DEBUG RAW TEXT:\n"
            out += repr(question_text[:500]) + "\n"
            out += "="*60 + "\n\n"
            
            # Parse question and answer choices
            import re
            # Look for answer choices pattern (A. or A) followed by text)
            lines = question_text.split('\n')
            question_part = []
            choices_part = []
            in_choices = False
            
            for line in lines:
                # Check if line starts with A-E followed by . or )
                if re.match(r'^[A-E][\.\)]', line.strip()):
                    in_choices = True
                    choices_part.append(line)
                elif in_choices:
                    # Continue collecting choices if they span multiple lines
                    if line.strip() and not re.match(r'^[A-E][\.\)]', line.strip()):
                        choices_part[-1] += " " + line.strip()
                    elif re.match(r'^[A-E][\.\)]', line.strip()):
                        choices_part.append(line)
                else:
                    question_part.append(line)
            
            # Display question
            out += '\n'.join(question_part).strip() + "\n\n"
            
            # Display choices if found
            if choices_part:
                out += "Answer Choices:\n"
                for choice in choices_part:
                    out += choice.strip() + "\n"
                out += "\n"
            
            out += f"Correct Answer: {r['metadatas'][0][i].get('answer', 'N/A')}\n"
            out += f"Similarity: {1 - r['distances'][0][i]:.3f}\n"
        return out
    except Exception as e:
        return f"Error: {e}"

demo = gr.Interface(
    fn=ui_search,
    inputs=[
        gr.Textbox(label="Medical Query", placeholder="e.g., hyponatremia"),
        gr.Slider(1, 5, value=3, step=1, label="Results")
    ],
    outputs=gr.Textbox(label="Similar USMLE Questions", lines=20),
    title="MedQA Search",
    examples=[["hyponatremia", 3], ["myocardial infarction", 2]]
)

# FastAPI
app = FastAPI()

class SearchRequest(BaseModel):
    query: str
    num_results: int = 3

@app.post("/search_medqa")
def api_search(req: SearchRequest):
    r = search(req.query, req.num_results)
    return {"results": [{
        "example_number": i+1,
        "question": r['documents'][0][i],
        "answer": r['metadatas'][0][i].get('answer', 'N/A'),
        "distance": r['distances'][0][i]
    } for i in range(len(r['documents'][0]))]}

app = gr.mount_gradio_app(app, demo, path="/")

# Launch the server
if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)