File size: 5,704 Bytes
8b442b3 bd2d247 8b442b3 |
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 |
import os
import pdfplumber
import docx
import pandas as pd
import numpy as np
import streamlit as st
from sentence_transformers import SentenceTransformer
from transformers import AutoTokenizer
import faiss
from groq import Groq
# ==========================================================
# GROQ API KEY (use HF Secrets)
# ==========================================================
os.environ["GROQ_API_KEY"] = os.getenv("API")
# ==========================================================
# STREAMLIT UI
# ==========================================================
st.set_page_config(page_title="Universal RAG App", layout="wide")
st.title("📄 Universal Document RAG (PDF | Word | Excel)")
uploaded_file = st.file_uploader(
"Upload a document",
type=["pdf", "docx", "xlsx"]
)
# ==========================================================
# TEXT EXTRACTION FUNCTIONS (UNCHANGED)
# ==========================================================
def read_pdf_with_plumber(pdf_path):
pages = []
with pdfplumber.open(pdf_path) as pdf:
for i, page in enumerate(pdf.pages):
text = page.extract_text(x_tolerance=2)
if text:
pages.append({"page": i + 1, "text": text})
return pages
def read_word(doc_path):
doc = docx.Document(doc_path)
text = "\n\n".join([p.text for p in doc.paragraphs if p.text.strip() != ""])
return [{"page": 1, "text": text}]
def read_excel(xlsx_path):
df = pd.read_excel(xlsx_path, sheet_name=None)
texts = []
for sheet_name, sheet in df.items():
sheet_text = sheet.fillna("").astype(str).agg(" ".join, axis=1).str.cat(sep="\n")
texts.append({"page": sheet_name, "text": sheet_text})
return texts
# ==========================================================
# CORE RAG FUNCTIONS (UNCHANGED)
# ==========================================================
def chunk_text(pages, chunk_size=800):
chunks = []
for page in pages:
paragraphs = page["text"].split("\n\n")
buffer = ""
for para in paragraphs:
if len(buffer) + len(para) <= chunk_size:
buffer += " " + para
else:
chunks.append({"page": page["page"], "text": buffer.strip()})
buffer = para
if buffer:
chunks.append({"page": page["page"], "text": buffer.strip()})
return chunks
def tokenize_chunks(chunks, model_name="sentence-transformers/all-mpnet-base-v2"):
tokenizer = AutoTokenizer.from_pretrained(model_name)
return [tokenizer(c["text"], truncation=True)["input_ids"] for c in chunks]
def create_embeddings(chunks, model_name="allenai/specter"):
embedder = SentenceTransformer(model_name)
texts = [c["text"] for c in chunks]
embeddings = embedder.encode(texts, show_progress_bar=False)
return embedder, np.array(embeddings)
def store_embeddings(embeddings):
faiss.normalize_L2(embeddings)
dim = embeddings.shape[1]
index = faiss.IndexFlatIP(dim)
index.add(embeddings)
return index
def retrieve_chunks(query, embedder, index, chunks, top_k=None):
if not top_k:
top_k = min(20, len(chunks))
query_vec = embedder.encode([query])
faiss.normalize_L2(query_vec)
scores, indices = index.search(query_vec, top_k)
return [chunks[i] for i in indices[0]]
def build_safe_context(retrieved_chunks, max_chars=12000):
context = ""
used = 0
for c in retrieved_chunks[:3]:
block = f"(Page {c['page']}) {c['text']}\n\n"
context += block
used += len(block)
for c in retrieved_chunks[3:]:
block = f"(Page {c['page']}) {c['text']}\n\n"
if used + len(block) > max_chars:
break
context += block
used += len(block)
return context
def generate_answer(query, context):
client = Groq()
prompt = f"""
You are a document-based assistant.
Use the context to answer the question clearly.
If the answer is partially available, summarize it.
If the answer is not present, you may say 'Not found in the document'.
Context:
{context}
Question:
{query}
"""
response = client.chat.completions.create(
model="llama-3.1-8b-instant",
messages=[{"role": "user", "content": prompt}],
temperature=0.3
)
return response.choices[0].message.content
# ==========================================================
# APP LOGIC
# ==========================================================
if uploaded_file:
with st.spinner("📄 Reading document..."):
file_name = uploaded_file.name
with open(file_name, "wb") as f:
f.write(uploaded_file.getbuffer())
if file_name.lower().endswith(".pdf"):
pages = read_pdf_with_plumber(file_name)
elif file_name.lower().endswith(".docx"):
pages = read_word(file_name)
elif file_name.lower().endswith(".xlsx"):
pages = read_excel(file_name)
else:
st.error("Unsupported file type")
with st.spinner("✂️ Chunking & embedding document..."):
chunks = chunk_text(pages)
tokenize_chunks(chunks)
embedder, embeddings = create_embeddings(chunks)
index = store_embeddings(embeddings)
st.success("✅ Document indexed successfully")
query = st.text_input("❓ Ask a question")
if query:
with st.spinner("🤖 Generating answer..."):
retrieved_chunks = retrieve_chunks(query, embedder, index, chunks)
context = build_safe_context(retrieved_chunks)
answer = generate_answer(query, context)
st.markdown("### ✅ Answer")
st.write(answer)
|