QuantCoder is a tool designed to streamline the process of searching for research articles, downloading PDFs, summarizing content, and generating QuantConnect Python algorithms based on the extracted data.
For an explanation of the code, refer to article : https://medium.com/ai-advances/from-finance-papers-to-trading-algorithms-an-automated-approach-ccd2180ee306?sk=c1e67131cd822bccc1acab1b53ae5331
This code is now integrated as coding engine in the project QuantCoder_FS, expected to be released Q2 2025. Refer to aricle : https://medium.com/ai-advances/towards-automating-quantitative-finance-research-c868a2a6477e Screenshots of the development are visible in a dedicated folder QuantCoder_FS_Demo.
- Search Articles: Query the CrossRef API to find relevant journal articles.
- List Articles: View previously searched articles.
- Download PDFs: Download article PDFs using direct links or Unpaywall.
- Summarize Articles: Generate concise summaries of downloaded articles.
- Generate QuantConnect Code: Create QuantConnect Python algorithms based on article summaries.
- Interactive Mode: Perform all steps interactively with guided prompts.
Detailed installation instructions will be provided once the CLI is fully set up.
To launch the interactive mode of QuantCLI, follow these steps:
-
Open a terminal window (bash shell).
-
Navigate to the directory where
quantcliis installed. -
Type the following command and press
Enter:quantcli interactive
The project was initiated in November 2023 with the goal of leveraging large language models (LLMs) within the LangChain framework to autonomously develop a trading algorithm. The cognitive architecture underpinning this system is comprehensively detailed in the article: https://medium.com/towardsdev/dual-agent-chatbots-and-expert-systems-design-25e2cba434e9
The trading strategy generated by the system is elaborated in this article: https://medium.com/coinmonks/how-to-outperform-the-market-fe151b944c77 Following a LinkedIn post announcing this innovative approach, which garnered approximately 10,000 impressions, I published several articles on AI-assisted pair-coding. This positive reception inspired me to integrate the research, summarization of quantitative finance articles, and algorithm coding into a unified workflow. More insights can be found in the article: https://medium.com/ai-advances/from-finance-papers-to-trading-algorithms-an-automated-approach-ccd2180ee306 which received significant attention in publication AI advances.
Subsequently, I released the initial proof of concept workflow (v0.1) on GitHub. The codebase was then refactored to adhere more closely to object-oriented programming (OOP) standards in version 0.2, and a user interface was introduced in version 0.3.
As of December 2024, I am developing the full-stack version of this tool (v0.3). The next phase involves enhancing code generation accuracy through an agent-based workflow using the CrewAI framework. I anticipate releasing the beta version to the quantitative finance community in Q3 2025.
This project is licensed under the MIT License. See the LICENSE file for details.