Vision: A search agent specialized in financce, business, accounting domains to assist users in information retrieval; providing the information sources to help users evaluate the quality of generated responses.
- A search agent for websites (Yahoo Finance, Bloomberg, XBRL International) and local files (SEC 10K, XBRL files (eXtensible Business Reporting Language)).
- A powerful information search-and-retrieval engine to quickly locate relevant financial information from various sources, such as websites, reports, filings, and databases.
- For generated responses, users can check the sources, ensuring reliability and accuracy.
- This is a demo of FinLLMs on the HuggingFace's Open Financial LLM Leaderboard.
NO Trading Suggestions!
Current Progress:
-
Snapshot of the search agent: drag, resize and minimize; Providing information on user's current page.
-
Checking sources, which are very important and help reduce misinformation.
-
User's webpage list. A list of URLs (and APIs in the future) that users set higher priority to retrieve information.
Important: You will need an API key for running the agent. Please ask the project leader for the key.
-
Open a terminal and navigate to the directory where you want to set up the project.
-
Clone the repository by running the following command: "git clone https://github.com/FlyM1ss/FinGPT_for_RCOS.git"
-
Install the necessary dependencies as shown in the requirements.txt file in the root path via "pip install -r requirements.txt"
-
Go to extensions on a Chrome-Based browser and select developer mode
-
Click "Load unpacked" on the top left corner. Navigate to the folder "FinGPT_For_RCOS/ChatBot-Fin/Extension-ChatBot-Fin/src", select and load it.
- Now create a new terminal, then go to FinGPT_For_RCOS/ChatBot-Fin and run the command "python manage.py runserver" or "python3 manage.py runserver" if using Python3. Wait for the server to start. This should take no longer than a couple seconds.
- Navigate to https://finance.yahoo.com/ or https://www.bloomberg.com/. The search agent should automatically load and scrape the homepage.
- Start chatting!
NOTE: There is an option under Settings window for enabling local RAG. It is a function under development and additional steps are required to get it working. If you want to try it out, please contact Felix Tian (FlyM1ss) on discord. I will update the README once it is ready.
Immediate Next Steps:
- Local file parsing's document
- SEC Filing's parsing (should be very similar to CDM files parsing)
- Make the installation easier.
- Users cannot re-open the agent once it is closed. Need to fix it.
Future Plans:
- Reduce response time and sources loading by improving datascraping.py.
- Build a dynamic database for users: personalized version.
- Fix context window. Need to add logic handling the context window for Ask and Advanced Ask.
- Make user-specified web lists more "intelligent", if there is any way for the agent to auto-navigate some websites.
- Start to promote the demo more and collect questions from users.
- Test different FinLLMs models like LLaMA3, on the HuggingFace's Open Financial LLM Leaderboard.
Disclaimer: We are sharing codes for academic purposes under the MIT education license. Nothing herein is financial advice, and NOT a recommendation to trade real money. Please use common sense and always first consult a professional before trading or investing.