Spaces:
Sleeping
Sleeping
File size: 4,050 Bytes
2deda75 c5baecb 2deda75 | 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 | import os
import gradio as gr
import numpy as np
import faiss
from groq import Groq
from pypdf import PdfReader
from sentence_transformers import SentenceTransformer
from langchain_text_splitters import RecursiveCharacterTextSplitter
# =====================================================
# Configuration
# =====================================================
RELEVANCE_THRESHOLD = 1.2 # lower = stricter relevance
# =====================================================
# Initialize Groq Client
# =====================================================
client = Groq(api_key=os.environ.get("RAG-GROQ"))
# =====================================================
# Load Embedding Model
# =====================================================
embedding_model = SentenceTransformer("all-MiniLM-L6-v2")
# =====================================================
# Global Vector Store
# =====================================================
vector_store = None
stored_chunks = []
# =====================================================
# PDF Processing Function
# =====================================================
def process_pdf(pdf_file):
global vector_store, stored_chunks
reader = PdfReader(pdf_file)
full_text = ""
for page in reader.pages:
if page.extract_text():
full_text += page.extract_text() + "\n"
splitter = RecursiveCharacterTextSplitter(
chunk_size=500,
chunk_overlap=100
)
chunks = splitter.split_text(full_text)
embeddings = embedding_model.encode(chunks)
dimension = embeddings.shape[1]
vector_store = faiss.IndexFlatL2(dimension)
vector_store.add(np.array(embeddings))
stored_chunks = chunks
return "✅ PDF processed successfully. You can now ask questions."
# =====================================================
# Question Answering Function
# =====================================================
def answer_question(question):
if vector_store is None:
return "⚠️ Please upload and process a PDF first."
question_embedding = embedding_model.encode([question])
distances, indices = vector_store.search(
np.array(question_embedding), k=3
)
avg_distance = distances[0].mean()
context = ""
for idx in indices[0]:
context += stored_chunks[idx] + "\n"
# Relevance feedback
if avg_distance > RELEVANCE_THRESHOLD:
relevance_note = (
"⚠️ **Note:** This question is not directly answered in the document.\n"
"The response below is based on loosely related context.\n\n"
)
else:
relevance_note = ""
prompt = f"""
You are an honest and careful AI assistant.
Instructions:
- Answer ONLY using the provided context.
- If the answer is not explicitly stated, say:
"This is not directly mentioned in the document, but based on related context..."
Context:
{context}
Question:
{question}
"""
response = client.chat.completions.create(
model="llama-3.3-70b-versatile",
messages=[
{"role": "user", "content": prompt}
]
)
return relevance_note + response.choices[0].message.content
# =====================================================
# Gradio UI
# =====================================================
with gr.Blocks() as app:
gr.Markdown("## 📄 RAG-based PDF Question Answering (Groq + FAISS)")
gr.Markdown(
"Upload a PDF and ask questions. "
"The system will clearly tell you if an answer is not directly mentioned."
)
pdf_file = gr.File(label="Upload PDF", file_types=[".pdf"])
process_btn = gr.Button("Process PDF")
status_box = gr.Textbox(label="Status", interactive=False)
question_box = gr.Textbox(label="Ask a Question")
answer_box = gr.Textbox(label="Answer", lines=8)
process_btn.click(
process_pdf,
inputs=pdf_file,
outputs=status_box
)
question_box.submit(
answer_question,
inputs=question_box,
outputs=answer_box
)
app.launch()
|