File size: 10,456 Bytes
85484cb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
"""
Medical RAG System
"""

import gradio as gr
import numpy as np
import json
import re
from typing import Optional
from datasets import load_dataset
from rank_bm25 import BM25Okapi
from sentence_transformers import SentenceTransformer
from groq import Groq

# 1. ЗАВАНТАЖЕННЯ ДАНИХ

print("Завантажуємо датасет medmcqa...")
dataset = load_dataset("medmcqa", split="validation[:500]", trust_remote_code=True)

# Формуємо документи з питань + пояснень
raw_docs = []
for item in dataset:
    explanation = item.get("exp") or ""
    question = item["question"]
    options = [item.get(f"op{k}", "") for k in ["a", "b", "c", "d"]]
    correct_key = ["a", "b", "c", "d"][item["cop"]]
    correct_answer = item.get(f"op{correct_key}", "")
    subject = item.get("subject_name", "")

    text = f"Question: {question}\n"
    text += f"Options: A) {options[0]}  B) {options[1]}  C) {options[2]}  D) {options[3]}\n"
    text += f"Answer: {correct_answer}\n"
    if explanation:
        text += f"Explanation: {explanation}"

    raw_docs.append({
        "text": text,
        "subject": subject,
        "question": question,
        "answer": correct_answer,
    })

print(f"Завантажено {len(raw_docs)} документів")

# 2. CHUNKING

def chunk_documents(docs, chunk_size=300, overlap=50):

    chunks = []
    for idx, doc in enumerate(docs):
        text = doc["text"]
        words = text.split()
        if len(words) <= chunk_size:
            chunks.append({
                "text": text,
                "source_id": idx,
                "subject": doc["subject"],
                "question": doc["question"],
                "answer": doc["answer"],
            })
        else:
            start = 0
            while start < len(words):
                end = min(start + chunk_size, len(words))
                chunk_text = " ".join(words[start:end])
                chunks.append({
                    "text": chunk_text,
                    "source_id": idx,
                    "subject": doc["subject"],
                    "question": doc["question"],
                    "answer": doc["answer"],
                })
                if end == len(words):
                    break
                start += chunk_size - overlap
    return chunks

chunks = chunk_documents(raw_docs)
print(f"Отримано {len(chunks)} чанків після chunking")

# 3. BM25 RETRIEVER

tokenized_corpus = [c["text"].lower().split() for c in chunks]
bm25 = BM25Okapi(tokenized_corpus)

def bm25_search(query: str, top_k: int = 5):
    tokenized_query = query.lower().split()
    scores = bm25.get_scores(tokenized_query)
    top_indices = np.argsort(scores)[::-1][:top_k]
    return [(chunks[i], float(scores[i])) for i in top_indices if scores[i] > 0]

# 4. SEMANTIC (DENSE) RETRIEVER

print("Завантажуємо модель для семантичного пошуку...")
embedder = SentenceTransformer("all-MiniLM-L6-v2")

print("Обчислюємо ембедінги для всіх чанків...")
chunk_texts = [c["text"] for c in chunks]
chunk_embeddings = embedder.encode(chunk_texts, batch_size=64, show_progress_bar=True, convert_to_numpy=True)
print("Ембедінги готові!")

def semantic_search(query: str, top_k: int = 5):
    query_emb = embedder.encode([query], convert_to_numpy=True)[0]
    norms = np.linalg.norm(chunk_embeddings, axis=1) * np.linalg.norm(query_emb)
    norms = np.where(norms == 0, 1e-9, norms)
    scores = chunk_embeddings @ query_emb / norms
    top_indices = np.argsort(scores)[::-1][:top_k]
    return [(chunks[i], float(scores[i])) for i in top_indices]

# 5. HYBRID RETRIEVER

def hybrid_search(query: str, top_k: int = 5, use_bm25: bool = True, use_semantic: bool = True):
    if not use_bm25 and not use_semantic:
        return []

    results = {}

    if use_bm25:
        bm25_results = bm25_search(query, top_k=top_k * 2)
        for rank, (chunk, score) in enumerate(bm25_results):
            key = chunk["text"][:80]
            results[key] = results.get(key, {"chunk": chunk, "score": 0})
            results[key]["score"] += 1 / (rank + 1)  # reciprocal rank fusion

    if use_semantic:
        sem_results = semantic_search(query, top_k=top_k * 2)
        for rank, (chunk, score) in enumerate(sem_results):
            key = chunk["text"][:80]
            results[key] = results.get(key, {"chunk": chunk, "score": 0})
            results[key]["score"] += 1 / (rank + 1)

    sorted_results = sorted(results.values(), key=lambda x: x["score"], reverse=True)
    return [(r["chunk"], r["score"]) for r in sorted_results[:top_k]]

# 6. LLM (GROQ)

def generate_answer(query: str, context_chunks: list, groq_api_key: str) -> str:
    client = Groq(api_key=groq_api_key)

    context_parts = []
    for i, (chunk, score) in enumerate(context_chunks, 1):
        context_parts.append(f"[{i}] {chunk['text']}")
    context = "\n\n".join(context_parts)

    system_prompt = """You are a helpful medical assistant. Answer the user's question based ONLY on the provided context.
- Cite sources using square brackets like [1], [2] when you use information from them.
- If the context doesn't contain enough information, say so honestly.
- Be concise and accurate.
- Always mention relevant medical details from the context."""

    user_prompt = f"""Context:
{context}

Question: {query}

Answer (with citations like [1], [2]):"""

    response = client.chat.completions.create(
        model="llama-3.3-70b-versatile",
        messages=[
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": user_prompt},
        ],
        temperature=0.2,
        max_tokens=600,
    )
    return response.choices[0].message.content

# 7. ГОЛОВНА ФУНКЦІЯ RAG

def rag_query(
    query: str,
    groq_api_key: str,
    use_bm25: bool,
    use_semantic: bool,
    top_k: int,
):
    if not query.strip():
        return "⚠️ Введіть запитання.", "", ""
    if not groq_api_key.strip():
        return "⚠️ Введіть Groq API ключ.", "", ""
    if not use_bm25 and not use_semantic:
        return "⚠️ Увімкніть хоча б один метод пошуку.", "", ""

    try:
        # Пошук релевантних чанків
        retrieved = hybrid_search(query, top_k=top_k, use_bm25=use_bm25, use_semantic=use_semantic)

        if not retrieved:
            return "Не знайдено релевантних документів.", "", ""

        # Генерація відповіді
        answer = generate_answer(query, retrieved, groq_api_key)

        sources_md = "### Джерела (чанки)\n\n"
        for i, (chunk, score) in enumerate(retrieved, 1):
            subj = chunk.get("subject", "—")
            sources_md += f"**[{i}]** *(Subject: {subj}, Score: {score:.4f})*\n\n"
            sources_md += f"```\n{chunk['text'][:400]}{'...' if len(chunk['text']) > 400 else ''}\n```\n\n"

        methods = []
        if use_bm25 and use_semantic:
            methods.append("Hybrid (BM25 + Semantic)")
        elif use_bm25:
            methods.append("BM25 (keyword)")
        else:
            methods.append("Semantic (dense)")

        info = f"**Метод пошуку:** {', '.join(methods)} | **Знайдено чанків:** {len(retrieved)}"

        return answer, sources_md, info

    except Exception as e:
        return f"Помилка: {str(e)}", "", ""

# ──────────────────────────────────────────────
# 8. GRADIO UI
# ──────────────────────────────────────────────

with gr.Blocks(title="Medical RAG System", theme=gr.themes.Soft()) as demo:
    gr.Markdown("""
    #Medical RAG System
    **Retrieval-Augmented Generation** для відповідей на медичні запитання.

    Система використовує датасет **medmcqa** (медичні питання з іспитів) та поєднує
    BM25 (пошук по ключових словах) і семантичний пошук для знаходження релевантних джерел.
    """)

    with gr.Row():
        with gr.Column(scale=2):
            query_input = gr.Textbox(
                label="Ваше запитання",
                placeholder="Напр.: What is the mechanism of action of aspirin?",
                lines=2,
            )
            api_key_input = gr.Textbox(
                label="Groq API Key",
                placeholder="gsk_...",
                type="password",
            )

        with gr.Column(scale=1):
            use_bm25 = gr.Checkbox(label="BM25 (keyword search)", value=True)
            use_semantic = gr.Checkbox(label="Semantic search", value=True)
            top_k = gr.Slider(minimum=1, maximum=10, value=5, step=1, label="Кількість чанків (top-k)")
            submit_btn = gr.Button("🔍 Знайти відповідь", variant="primary")

    info_box = gr.Markdown("")

    with gr.Tabs():
        with gr.Tab("Відповідь"):
            answer_output = gr.Markdown(label="Відповідь")
        with gr.Tab("Джерела"):
            sources_output = gr.Markdown(label="Використані чанки")

    gr.Examples(
        examples=[
            ["What is the mechanism of action of aspirin?", True, True, 5],
            ["Which vitamin deficiency causes night blindness?", True, True, 5],
            ["What are symptoms of diabetes mellitus?", False, True, 5],
            ["beta blocker mechanism", True, False, 5],
        ],
        inputs=[query_input, use_bm25, use_semantic, top_k],
        label="Приклади запитів",
    )

    submit_btn.click(
        fn=rag_query,
        inputs=[query_input, api_key_input, use_bm25, use_semantic, top_k],
        outputs=[answer_output, sources_output, info_box],
    )

    gr.Markdown("""
    ---
    **Датасет:** [medmcqa](https://huggingface.co/datasets/medmcqa) |
    **LLM:** Groq (llama3-8b-8192) |
    **Embeddings:** all-MiniLM-L6-v2 |
    **Chunking:** sliding window (300 слів, overlap 50)
    """)

if __name__ == "__main__":
    demo.launch(share=True)