summaryrefslogtreecommitdiff
path: root/rag/ui.py
blob: 5a5e01aa1176685b0833fc7d56f5313a8e614ea3 (plain)
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
from typing import Type

import streamlit as st
from dotenv import load_dotenv
from langchain_community.document_loaders.blob_loaders import Blob

from rag.generator import MODELS, get_generator
from rag.generator.abstract import AbstractGenerator
from rag.generator.prompt import Prompt
from rag.retriever.retriever import Retriever


@st.cache_resource
def load_retriever() -> Retriever:
    return Retriever()


@st.cache_resource
def load_generator(model: str) -> Type[AbstractGenerator]:
    return get_generator(model)


@st.cache_data(show_spinner=False)
def upload(files):
    with st.spinner("Indexing documents..."):
        for file in files:
            source = file.name
            blob = Blob.from_data(file.read())
            retriever.add_pdf(blob=blob, source=source)


if __name__ == "__main__":
    load_dotenv()
    retriever = load_retriever()
    ss = st.session_state
    st.title("Retrieval Augmented Generation")

    model = st.selectbox("Generative Model", options=MODELS)
    generator = load_generator(model)

    files = st.file_uploader(
        "Choose pdfs to add to the knowledge base",
        type="pdf",
        accept_multiple_files=True,
    )

    upload(files)

    with st.form(key="query"):
        query = st.text_area(
            "query",
            key="query",
            height=100,
            placeholder="Enter query here",
            help="",
            label_visibility="collapsed",
            disabled=False,
        )
        submit = st.form_submit_button("Generate")

    (result_column, context_column) = st.columns(2)

    if submit:
        if not query:
            st.stop()

        query = ss.get("query", "")
        with st.spinner("Searching for documents..."):
            documents = retriever.retrieve(query)

        prompt = Prompt(query, documents)

        with context_column:
            st.markdown("### Context")
            for i, doc in enumerate(documents):
                st.markdown(f"### Document {i}")
                st.markdown(f"**Title: {doc.title}**")
                st.markdown(doc.text)
                st.markdown("---")

        with result_column:
            st.markdown("### Answer")
            st.write_stream(generator.generate(prompt))