Conference Assistant

🤖 create | 💫 chat | 💥 enhance | 🤝 customize | 🏷 license

English | 中文
A conference assistant based on the LLMs, which can be used for various types of conferences such as company annual meetings and academic conferences. It is very easy to use, just provide the meeting documents.
Include but are not limited to the following purposes:

• Basic information of the meeting
• Specific agenda
• Hotel and itinerary arrangement
• Other precautions

🤖 create your exclusive conference assistant

You need to specify the name of the company, school, or organization, and you can also add some instructions to restrict or optimize the assistant's responses.

from assistant import LLMAssistant

# Assistant Settings.
firm = 'CITIC Guoan Industrial Group Co., Ltd' 
instruction = '' 
model = 'qwen-max'

# Conference Information.
file_path = ['./data/agenda_example.xlsx'] 
        
# Create Conference Assistant.
my_assistant = LLMAssistant(firm=firm, instruction=instrction, file_path=file_path, model=model)

You need to specify the large language model used by the assistant. We follow the OpenAI SDK, and currently support the following models: qwen、kimi、spark. If there are many documents, it is recommended to use a model that supports longer context input.

# Add your api_key to the environment variables.
import os
os.environ['API_KEY'] = 'sk-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX'

# Support the following models:
# qwen: qwen-plus, qwen-max  
# kimi: moonshot-v1-8k, moonshot-v1-32k  
# spark: generalv3.5  

You need to provide additional conference documents, and the assistant will learn and answer based on the content of these documents. We currently support a variety of file formats, including .doc, .docx, .xls, .xlsx, .txt, .csv, .tsv, and more. However, we recommend using .xlsx or .csv formats for optimal results.

# Check the conference content used by the assistant.
content = my_assistant.conference_info()
print('conference content: ', content)

💫 quickly chat with your assistant

Quickly initiate chat, support streaming and non streaming, default streaming.

# Start a Chat.
query = 'What should I call you?'
stream = False

answer = ''
for chunk in my_assistant.single_chat(query=query, stream=stream):
    answer += chunk
print('answer:', answer)

💥 start a more powerful chat

By combining historical session information, the assistant can fully understand each query and answer more accurately.
When chatting, you need to specify a session ID to manage session information and specify the session rounds that the assistant should consider when answering.

# Start a Chat.
session_id = '20250108001'
query = 'What should I call you?'
stream = True
rounds = 3

answer = ''
for chunk in my_assistant.chat(session_id=session_id, query=query, stream=stream, rounds=rounds):
    answer += chunk
print('answer:', answer)

Multiple rounds of session require middleware to store and manage session information. Please change your Redis configuration in the config.py as shown below:

redis_config = {
    'host':'127.0.0.1',
    'port':'7001',
    'password':'xxxxxxxxx'
}

Manage historical information through session ID.

history = my_assistant.history_info(session_id)
print('session history: ', history)

🤝 personalized customization

You can also explore additional customized developments triggered through natural language. If you have any needs in this area, please feel welcome to contact us:

• email : zhenhu317@gmail.com

🏷 license

Conference Assistant is licensed under the Apache-2.0 License.