Upload 9 files

Data/cancer_and_cure__a_critical_analysis.27.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcd5a00bfc0fdc26be5fac6be1e1e7a9a1a5fde56ad6238597fbf23168238e48
+size 225771

Data/medical_oncology_handbook_june_2020_edition.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e434ce92bd5b21c9da017f476535359dd400858d78083c89a817bea0d445abd
+size 818286

README.md

@@ -1,11 +1,2 @@
----
-title: Qdrant Backend
-emoji: 👁
-colorFrom: indigo
-colorTo: yellow
-sdk: docker
-pinned: false
-license: mit
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Medical RAG-using-Meditron-7B-LLM
+Medical RAG QA App using Meditron 7B LLM, Qdrant Vector Database, and PubMedBERT Embedding Model.

app.py

@@ -0,0 +1,80 @@
+from langchain import PromptTemplate
+from langchain.llms import CTransformers
+from langchain.chains import RetrievalQA
+from langchain.embeddings import SentenceTransformerEmbeddings
+from fastapi import FastAPI, Request, Form, Response
+from fastapi.responses import HTMLResponse
+from fastapi.templating import Jinja2Templates
+from fastapi.staticfiles import StaticFiles
+from fastapi.encoders import jsonable_encoder
+from qdrant_client import QdrantClient
+from langchain.vectorstores import Qdrant
+import os
+import json
+
+app = FastAPI()
+templates = Jinja2Templates(directory="templates")
+app.mount("/static", StaticFiles(directory="static"), name="static")
+
+local_llm = "joshnader/meditron-7b-Q4_K_M-GGUF"
+
+config = {
+'max_new_tokens': 512,
+'context_length': 2048,
+'repetition_penalty': 1.1,
+'temperature': 0.1,
+'top_k': 50,
+'top_p': 0.9,
+'stream': True,
+'threads': int(os.cpu_count() / 4)
+}
+
+llm = CTransformers(
+    model=local_llm,
+    model_type="llama",
+    **config
+)
+
+print("LLM Initialized....")
+
+prompt_template = """Use the following pieces of information to answer the user's question.
+If you don't know the answer, just say that you don't know, don't try to make up an answer.
+
+Context: {context}
+Question: {question}
+
+Only return the helpful answer below and nothing else.
+Helpful answer:
+"""
+
+embeddings = SentenceTransformerEmbeddings(model_name="NeuML/pubmedbert-base-embeddings")
+
+client = QdrantClient(
+    url=os.getenv("QDRANT_URL", "https://868005ec-814c-4a06-b5f5-f4051fdf2a5d.europe-west3-0.gcp.cloud.qdrant.io"),
+    api_key=os.getenv("QDRANT_API_KEY"),
+    prefer_grpc=False
+)
+
+db = Qdrant(client=client, embeddings=embeddings, collection_name="vector_db")
+
+prompt = PromptTemplate(template=prompt_template, input_variables=['context', 'question'])
+
+retriever = db.as_retriever(search_kwargs={"k":1})
+
+@app.get("/", response_class=HTMLResponse)
+async def read_root(request: Request):
+    return templates.TemplateResponse("index.html", {"request": request})
+
+@app.post("/get_response")
+async def get_response(query: str = Form(...)):
+    chain_type_kwargs = {"prompt": prompt}
+    qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=retriever, return_source_documents=True, chain_type_kwargs=chain_type_kwargs, verbose=True)
+    response = qa(query)
+    print(response)
+    answer = response['result']
+    source_document = response['source_documents'][0].page_content
+    doc = response['source_documents'][0].metadata['source']
+    response_data = jsonable_encoder(json.dumps({"answer": answer, "source_document": source_document, "doc": doc}))
+    
+    res = Response(response_data)
+    return res

rag.py

@@ -0,0 +1,83 @@
+from langchain import PromptTemplate
+from langchain.llms import CTransformers
+from langchain.chains import RetrievalQA
+from langchain.embeddings import SentenceTransformerEmbeddings
+from fastapi import FastAPI, Request, Form, Response
+from fastapi.responses import HTMLResponse
+from fastapi.templating import Jinja2Templates
+from fastapi.staticfiles import StaticFiles
+from fastapi.encoders import jsonable_encoder
+from qdrant_client import QdrantClient
+from langchain.vectorstores import Qdrant
+import os
+import json
+
+app = FastAPI()
+templates = Jinja2Templates(directory="templates")
+app.mount("/static", StaticFiles(directory="static"), name="static")
+
+local_llm = "joshnader/meditron-7b-Q4_K_M-GGUF"
+
+config = {
+'max_new_tokens': 512,
+'context_length': 2048,
+'repetition_penalty': 1.1,
+'temperature': 0.1,
+'top_k': 50,
+'top_p': 0.9,
+'stream': True,
+'threads': int(os.cpu_count() / 4)
+}
+
+
+llm = CTransformers(
+    model=local_llm,
+    model_type="llama",
+    **config
+)
+
+
+
+print("LLM Initialized....")
+
+prompt_template = """Use the following pieces of information to answer the user's question.
+If you don't know the answer, just say that you don't know, don't try to make up an answer.
+
+Context: {context}
+Question: {question}
+
+Only return the helpful answer below and nothing else.
+Helpful answer:
+"""
+
+embeddings = SentenceTransformerEmbeddings(model_name="NeuML/pubmedbert-base-embeddings")
+
+url = "http://localhost:6333"
+
+client = QdrantClient(
+    url=url, prefer_grpc=False
+)
+
+db = Qdrant(client=client, embeddings=embeddings, collection_name="vector_db")
+
+prompt = PromptTemplate(template=prompt_template, input_variables=['context', 'question'])
+
+retriever = db.as_retriever(search_kwargs={"k":1})
+
+@app.get("/", response_class=HTMLResponse)
+async def read_root(request: Request):
+    return templates.TemplateResponse("index.html", {"request": request})
+
+@app.post("/get_response")
+async def get_response(query: str = Form(...)):
+    chain_type_kwargs = {"prompt": prompt}
+    qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=retriever, return_source_documents=True, chain_type_kwargs=chain_type_kwargs, verbose=True)
+    response = qa(query)
+    print(response)
+    answer = response['result']
+    source_document = response['source_documents'][0].page_content
+    doc = response['source_documents'][0].metadata['source']
+    response_data = jsonable_encoder(json.dumps({"answer": answer, "source_document": source_document, "doc": doc}))
+    
+    res = Response(response_data)
+    return res

retriever.py

@@ -0,0 +1,25 @@
+from langchain.vectorstores import Qdrant
+from langchain.embeddings import SentenceTransformerEmbeddings
+from qdrant_client import QdrantClient
+
+embeddings = SentenceTransformerEmbeddings(model_name="NeuML/pubmedbert-base-embeddings")
+
+url = "http://localhost:6333"
+
+client = QdrantClient(
+    url=url, prefer_grpc=False
+)
+
+print(client)
+print("##############")
+
+db = Qdrant(client=client, embeddings=embeddings, collection_name="vector_db")
+
+print(db)
+print("######")
+query = "What is Metastatic disease?"
+
+docs = db.similarity_search_with_score(query=query, k=3)
+for i in docs:
+    doc, score = i
+    print({"score": score, "content": doc.page_content, "metadata": doc.metadata})

start.sh

@@ -0,0 +1,8 @@
+
+#!/bin/bash
+
+# Ingest PDFs into the vector database (only needed once)
+python ingest.py
+
+# Start the FastAPI server
+uvicorn app:app --host 0.0.0.0 --port 7860

templates/.gitattributes

@@ -0,0 +1 @@
+data/**/*.pdf filter=lfs diff=lfs merge=lfs -text

templates/index.html

@@ -0,0 +1,120 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>Oncology RAG App</title>
+    <link href="https://fonts.googleapis.com/css2?family=Poppins:wght@400;600&display=swap" rel="stylesheet">
+    <link href="https://cdn.jsdelivr.net/npm/[email protected]/dist/css/bootstrap.min.css" rel="stylesheet">
+    <style>
+        body {
+            background-color: black;
+            font-family: 'Poppins', sans-serif;
+            color: white;
+        }
+        .chat-container {
+            max-width: 800px;
+            margin: 50px auto;
+            margin-top: 10%;
+            padding: 20px;
+            background-color: #333;
+            border-radius: 10px;
+        }
+        .chat-heading {
+            text-align: center;
+            font-size: 2.5em;
+            font-weight: 600;
+            margin-bottom: 30px;
+            color: #ffd700; /* Golden color for the heading */
+        }
+        .chat-input {
+            margin-top: 20px; /* Added margin */
+            margin-bottom: 20px;
+            height: 100px; /* Increased height */
+        }
+        .chat-button {
+            background-color: green;
+            color: white;
+            padding: 10px 20px;
+            font-size: 1.2em;
+        }
+        .chat-response {
+            background-color: #444;
+            padding: 15px;
+            border-radius: 5px;
+            min-height: 100px; /* Minimum height for the response box */
+            margin-top: 20px;
+        }
+        .accordion {
+            margin-top: 20px;
+            background-color: #444;
+            border-radius: 5px;
+        }
+        .accordion-button {
+            color: white;
+            background-color: #555;
+        }
+        .accordion-body {
+            color: white; /* Improved visibility of text */
+        }
+        pre {
+            white-space:pre-wrap;
+        }
+    </style>
+</head>
+<body>
+    <div class="container chat-container">
+        <h1 class="chat-heading">Medical RAG QA App</h1>
+
+        <div class="accordion" id="appDescriptionAccordion">
+            <div class="accordion-item">
+                <h2 class="accordion-header" id="descriptionHeading">
+                    <button class="accordion-button collapsed" type="button" data-bs-toggle="collapse" data-bs-target="#collapseDescription" aria-expanded="true" aria-controls="collapseDescription">
+                        About This App
+                    </button>
+                </h2>
+                <div id="collapseDescription" class="accordion-collapse collapse" aria-labelledby="descriptionHeading" data-bs-parent="#appDescriptionAccordion">
+                    <div class="accordion-body text-dark">
+                        This is a RAG implementation using Open Source stack. Intel's Neural Chat has been used to build this app along with BGE Embeddings as an embedding model, Chroma DB as a vector store, and Langchain & CTransformers as an orchestration frameworks.
+                    </div>
+                </div>
+            </div>
+        </div>
+
+        <div class="row">
+            <div class="col">
+                <textarea id="userInput" class="form-control chat-input" placeholder="Type your query here..."></textarea>
+                <button id="submitBtn" class="btn chat-button">Submit</button>
+                <div id="response" class="chat-response"></div>
+            </div>
+        </div>
+    </div>
+
+    <script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/js/bootstrap.bundle.min.js"></script>
+    <script>
+        document.getElementById('submitBtn').addEventListener('click', async function() {
+            var userInput = document.getElementById('userInput').value;
+            document.getElementById('response').innerHTML = '<p>Processing...</p>';
+            const formData = new FormData();
+            formData.append('query', userInput); 
+            try {
+                const response = await fetch('/get_response', {
+                    method: 'POST',
+                    body: formData
+                });
+
+                if (!response.ok) {
+                    throw new Error('Network response was not ok');
+                }
+
+                const data = await response.json();
+                document.getElementById('response').innerHTML = `<p>${data.answer}</p><br><pre><b>Context: </b> ${data.source_document}</pre><br><pre><b>Source Document: </b> ${data.doc}</pre>`;
+            } catch (error) {
+                console.error('Error:', error);
+                document.getElementById('response').innerHTML = '<p>Error processing your request</p>';
+            }
+        });
+
+    </script>
+</body>
+</html>