my-project-public

repository

loading code, commits, and activity

repositories

loading repo index

#1	import os
#2	import queue
#3	import re
#4	import tempfile
#5	import threading
#6
#7	import streamlit as st
#8
#9	from embedchain import App
#10	from embedchain.config import BaseLlmConfig
#11	from embedchain.helpers.callbacks import StreamingStdOutCallbackHandlerYield, generate
#12
#13
#14	def embedchain_bot(db_path, api_key):
#15	return App.from_config(
#16	config={
#17	"llm": {
#18	"provider": "openai",
#19	"config": {
#20	"model": "gpt-4o-mini",
#21	"temperature": 0.5,
#22	"max_tokens": 1000,
#23	"top_p": 1,
#24	"stream": True,
#25	"api_key": api_key,
#26	},
#27	},
#28	"vectordb": {
#29	"provider": "chroma",
#30	"config": {"collection_name": "chat-pdf", "dir": db_path, "allow_reset": True},
#31	},
#32	"embedder": {"provider": "openai", "config": {"api_key": api_key}},
#33	"chunker": {"chunk_size": 2000, "chunk_overlap": 0, "length_function": "len"},
#34	}
#35	)
#36
#37
#38	def get_db_path():
#39	tmpdirname = tempfile.mkdtemp()
#40	return tmpdirname
#41
#42
#43	def get_ec_app(api_key):
#44	if "app" in st.session_state:
#45	print("Found app in session state")
#46	app = st.session_state.app
#47	else:
#48	print("Creating app")
#49	db_path = get_db_path()
#50	app = embedchain_bot(db_path, api_key)
#51	st.session_state.app = app
#52	return app
#53
#54
#55	with st.sidebar:
#56	openai_access_token = st.text_input("OpenAI API Key", key="api_key", type="password")
#57	"WE DO NOT STORE YOUR OPENAI KEY."
#58	"Just paste your OpenAI API key here and we'll use it to power the chatbot. [Get your OpenAI API key](https://platform.openai.com/api-keys)" # noqa: E501
#59
#60	if st.session_state.api_key:
#61	app = get_ec_app(st.session_state.api_key)
#62
#63	pdf_files = st.file_uploader("Upload your PDF files", accept_multiple_files=True, type="pdf")
#64	add_pdf_files = st.session_state.get("add_pdf_files", [])
#65	for pdf_file in pdf_files:
#66	file_name = pdf_file.name
#67	if file_name in add_pdf_files:
#68	continue
#69	try:
#70	if not st.session_state.api_key:
#71	st.error("Please enter your OpenAI API Key")
#72	st.stop()
#73	temp_file_name = None
#74	with tempfile.NamedTemporaryFile(mode="wb", delete=False, prefix=file_name, suffix=".pdf") as f:
#75	f.write(pdf_file.getvalue())
#76	temp_file_name = f.name
#77	if temp_file_name:
#78	st.markdown(f"Adding {file_name} to knowledge base...")
#79	app.add(temp_file_name, data_type="pdf_file")
#80	st.markdown("")
#81	add_pdf_files.append(file_name)
#82	os.remove(temp_file_name)
#83	st.session_state.messages.append({"role": "assistant", "content": f"Added {file_name} to knowledge base!"})
#84	except Exception as e:
#85	st.error(f"Error adding {file_name} to knowledge base: {e}")
#86	st.stop()
#87	st.session_state["add_pdf_files"] = add_pdf_files
#88
#89	st.title("📄 Embedchain - Chat with PDF")
#90	styled_caption = '<p style="font-size: 17px; color: #aaa;">🚀 An <a href="https://github.com/embedchain/embedchain">Embedchain</a> app powered by OpenAI!</p>' # noqa: E501
#91	st.markdown(styled_caption, unsafe_allow_html=True)
#92
#93	if "messages" not in st.session_state:
#94	st.session_state.messages = [
#95	{
#96	"role": "assistant",
#97	"content": """
#98	Hi! I'm chatbot powered by Embedchain, which can answer questions about your pdf documents.\n
#99	Upload your pdf documents here and I'll answer your questions about them!
#100	""",
#101	}
#102	]
#103
#104	for message in st.session_state.messages:
#105	with st.chat_message(message["role"]):
#106	st.markdown(message["content"])
#107
#108	if prompt := st.chat_input("Ask me anything!"):
#109	if not st.session_state.api_key:
#110	st.error("Please enter your OpenAI API Key", icon="🤖")
#111	st.stop()
#112
#113	app = get_ec_app(st.session_state.api_key)
#114
#115	with st.chat_message("user"):
#116	st.session_state.messages.append({"role": "user", "content": prompt})
#117	st.markdown(prompt)
#118
#119	with st.chat_message("assistant"):
#120	msg_placeholder = st.empty()
#121	msg_placeholder.markdown("Thinking...")
#122	full_response = ""
#123
#124	q = queue.Queue()
#125
#126	def app_response(result):
#127	llm_config = app.llm.config.as_dict()
#128	llm_config["callbacks"] = [StreamingStdOutCallbackHandlerYield(q=q)]
#129	config = BaseLlmConfig(**llm_config)
#130	answer, citations = app.chat(prompt, config=config, citations=True)
#131	result["answer"] = answer
#132	result["citations"] = citations
#133
#134	results = {}
#135	thread = threading.Thread(target=app_response, args=(results,))
#136	thread.start()
#137
#138	for answer_chunk in generate(q):
#139	full_response += answer_chunk
#140	msg_placeholder.markdown(full_response)
#141
#142	thread.join()
#143	answer, citations = results["answer"], results["citations"]
#144	if citations:
#145	full_response += "\n\nSources:\n"
#146	sources = []
#147	for i, citation in enumerate(citations):
#148	source = citation[1]["url"]
#149	pattern = re.compile(r"([^/]+)\.[^\.]+\.pdf$")
#150	match = pattern.search(source)
#151	if match:
#152	source = match.group(1) + ".pdf"
#153	sources.append(source)
#154	sources = list(set(sources))
#155	for source in sources:
#156	full_response += f"- {source}\n"
#157
#158	msg_placeholder.markdown(full_response)
#159	print("Answer: ", full_response)
#160	st.session_state.messages.append({"role": "assistant", "content": full_response})
#161

z6Mkq5mY3JWtxoxUobWcfNHm7AkRubgSWEZTkBVqZXJviFZ5/my-project-public