Spaces:
Sleeping
Sleeping
File size: 6,353 Bytes
79e9767 |
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 |
"""
Streamlit app
"""
try:
__import__('pysqlite3')
import sys
sys.modules['sqlite3'] = sys.modules.pop('pysqlite3')
except:
pass
import os
import streamlit as st
from ragvizexpander import RAGVizChain
from ragvizexpander.llms import *
from ragvizexpander.embeddings import *
from ragvizexpander.splitters import RecursiveChar2TokenSplitter
st.set_page_config(
page_title="RAGVizExpander Demo",
page_icon="🔬",
layout="wide"
)
os.environ['OPENAI_API_KEY'] = st.secrets["OPENAI_API_KEY"]
os.environ['HF_API_KEY'] = st.secrets["HF_API_KEY"]
if "chart" not in st.session_state:
st.session_state['chart'] = None
if "loaded" not in st.session_state:
st.session_state['loaded'] = False
st.title("RAGVizExpander Demo🔬")
st.markdown("📦 More details can be found at the GitHub repo [here](https://github.com/KKenny0/RAGVizExpander)")
if not st.session_state['loaded']:
main_page = st.empty()
main_button = st.empty()
with main_page.container():
uploaded_file = st.file_uploader("Upload your file",
label_visibility="collapsed",
type=['pdf', 'docx', 'txt', 'pptx'])
# --- setting llm model
st.markdown("### Settings for *LLM* model")
st.session_state["llm_model_type"] = st.radio("Select type of llm model",
["OpenAI", "Ollama"],
horizontal=True)
if st.session_state["llm_model_type"] == "OpenAI":
st.session_state["openai_llm_base_url"] = st.text_input("Enter OpenAI LLM API Base")
st.session_state["openai_llm_api_key"] = st.text_input("Enter OpenAI LLM API Key")
st.session_state["openai_llm_model"] = st.text_input("Enter OpenAI LLM model name")
st.session_state["chosen_llm_model"] = ChatOpenAI(
base_url=st.session_state["openai_llm_base_url"],
api_key=st.session_state["openai_llm_api_key"],
model_name=st.session_state["openai_llm_model"],
)
else:
st.session_state["ollama_llm_model"] = st.text_input("Enter Ollama model name")
st.session_state["chosen_llm_model"] = ChatOllama(model_name=st.session_state["ollama_llm_model"])
st.markdown("""---""")
# --- setting embedding model
st.markdown("### Settings for *EMBEDDING* model")
st.session_state["embedding_model_type"] = st.radio("Select type of embedding model",
["OpenAI", "SentenceTransformer", "HuggingFace", "TEI"],
horizontal=True)
if st.session_state["embedding_model_type"] == "OpenAI":
st.session_state["openai_embed_model"] = st.selectbox("Select embedding model",
["text-embedding-3-small",
"text-embedding-3-large",
"text-embedding-ada-002"])
st.session_state["openai_embed_api_key"] = st.text_input("Enter OpenAI Embedding API Key")
st.session_state["openai_embed_api_base"] = st.text_input("Enter OpenAI Embedding API Base")
st.session_state["chosen_embedding_model"] = OpenAIEmbeddings(
api_base=st.session_state["openai_embed_api_base"],
api_key=st.session_state["openai_embed_api_key"],
model_name=st.session_state["openai_embed_model"],
)
elif st.session_state["embedding_model_type"] == "HuggingFace":
st.session_state["hf_embed_model"] = st.text_input("Enter HF repository name")
st.session_state["hf_api_key"] = st.text_input("Enter HF API key")
st.session_state["chosen_embedding_model"] = HuggingFaceEmbeddings(
model_name=st.session_state["hf_embed_model"],
api_key=st.session_state["hf_api_key"]
)
else:
st.session_state["tei_api_url"] = st.text_input("Enter TEI(Text-Embedding-Inference) api url")
st.session_state["chosen_embedding_model"] = TEIEmbeddings(
api_url=st.session_state["tei_api_url"]
)
st.markdown("""---""")
# --- setting chunking parameters
st.markdown("### Settings for *CHUNKING* model")
st.session_state["chunk_size"] = st.number_input("Chunk size", value=500, min_value=100, max_value=1000, step=100)
st.session_state["chunk_overlap"] = st.number_input("Chunk overlap", value=0, min_value=0, max_value=100, step=10)
st.session_state["split_func"] = RecursiveChar2TokenSplitter(
chunk_size=st.session_state["chunk_size"],
chunk_overlap=st.session_state["chunk_overlap"],
)
if st.button("Build Vector DB"):
st.session_state["client"] = RAGVizChain(embedding_model=st.session_state["chosen_embedding_model"],
llm=st.session_state["chosen_llm_model"],
split_func=st.session_state["split_func"])
main_page.empty()
main_button.empty()
with st.spinner("Building Vector DB"):
st.session_state["client"].load_data(uploaded_file,)
st.session_state['loaded'] = True
st.rerun()
else:
col1, col2 = st.columns(2)
st.session_state['query'] = col1.text_area("Enter your query here")
st.session_state['technique'] = col1.radio("Select retrival technique", ["naive", "HyAE", "multi_qns"], horizontal=True)
st.session_state['top_k'] = col1.number_input("Top k", value=5, min_value=1, max_value=10, step=1)
if col1.button("Execute Query"):
st.session_state['chart'] = st.session_state["client"].visualize_query(st.session_state['query'], retrieval_method=st.session_state['technique'], top_k=st.session_state['top_k'])
if st.session_state['chart'] is not None:
col2.plotly_chart(st.session_state['chart'])
if col1.button("Reset Application"):
st.session_state['loaded'] = False
st.rerun()
|