Financial Agent

Overview

This repository contains a suite of Python scripts designed to automate the downloading, processing, and cleaning of financial filings from the U.S. Securities and Exchange Commission (SEC) database. These tools are intended for anyone interested in performing detailed analyses of corporate financial statements, especially through the use of Large language models.

Scripts Included

• SEC Filings Downloader (sec-filings-downloader.py): Downloads XBRL and TXT filings using company CIKs.
• SEC Metadata Processor (sec-metadata-processor.py): Extracts and processes financial statements and metadata from filings.
• HTML Filings Cleaner (financial_filings_htm_cleaner.py): Cleans and prepares HTM financial filings for analysis.
• Custom HTML Document Loader and Chunking (custom_html_loader_and_chunking.ipynb): Load the cleaned HTM financial filings and then perform chunking

Features

• Download filings for multiple companies.
• Handles both XBRL (zipped) and plain text formats.
• Filters filings by form type (e.g., 10-K, 10-Q) and date range.
• Document loader and chunking: refer to the notebook "custom_html_loader_and_chunking.ipynb"

Usage

1. Clone this repository to your local machine.
2. Modify the BASE_DIR in the scripts to the path where you want to save the filings.
3. Optionally, update the COMPANY_CIKS dictionary in each script to include the CIKs of the companies you're interested in.
4. Run the desired script using Python 3.11.

Quick Start

python sec-filings-downloader.py # Download filings python sec-metadata-processor.py # Process and extract data python financial_filings_htm_cleaner.py # Clean HTML filings

Requirements

python = "^3.11" langchain = "^0.1.11" langchain-openai = "^0.0.8" langchain-core = "^0.1.29" langchain-community = "^0.0.26" fastapi = "^0.110.0" python-dotenv = "^1.0.1" langchain-anthropic = "^0.1.3" beautifulsoup4 = "^4.12.3" lark = "^1.1.9" chromadb = "^0.4.24" langchain-elasticsearch = "^0.1.2"

Contributing

Contributions to financial_agent are welcome! Here are some ways you can contribute:

Submit pull requests with improvements to the scripts or documentation. Report any issues you encounter using the scripts. Suggest new features or enhancements. Feel free to fork this repository and submit your contributions via pull requests.

License

This project is open source under the MIT license. See the LICENSE file for more information.