Pssst, You! The original story is on Medium! Please open the article, clap a few times, and forward this email to your friend interested in AI!
|
Two months ago, I was trying to fix a bug in my AI-Powered automated investing platform, NexusTrade. Specifically, the issue was with the “Create a Portfolio Prompt” feature, which is designed to generate a unique set of trading strategies based on user inputs. It was frustrating trying to perform the prompt engineering in the source code, and it was difficult to isolate and test the functionality.
|
At the same time, I had a sorta revelation: the AI-Powered Chat is NexusTrade’s most popular feature by far. It allowed the configuration of highly sophisticated trading strategies, even for people without a technical background. The core principles of the chat — automating complex processes and improving the user experience — can be applied to a wide range of applications. For example, imagine if a teacher could generate a highly customizable lesson plan for their students that’s tailored depending on each student’s interests and personality. If only they had a platform that made this easy.
|
That’s why I set out to build NexusGenAI. NexusGenAI branches off from NexusTrade and abstracts the AI-Powered Chat to a separate application. At the same time, it allows the configuration of Large Language Model (LLM)-Powered applications in an easy-to-use user interface.
|
This article will describe the current functionality of NexusGenAI and the migration of NexusTrade’s Chat to the platform. It will then describe the next steps before the application is ready for a large-scale public release.
|
If you want to be the first to try NexusGenAI, sign up for the waitlist today!
|
Transitioning NexusTrade’s AI Chat to NexusGenAI
|
The initial step in developing NexusGenAI involves establishing a minimum viable product (MVP). The primary goal is to integrate the sophisticated capabilities of NexusTrade’s chat into this new application. This will encompass features such as the creation and backtesting of trading strategies, along with advanced financial analysis. For a comprehensive understanding of NexusTrade’s chat architecture, refer to the detailed description provided in the following article:
|
To summarize the article, the NexusTrade application operates through a sequential process:
|
- It identifies the user’s request as pertaining to a specific “Prompt.”
- It utilizes the corresponding Prompt to formulate a response to the user.
- Optionally, it engages in post-processing steps, which may include executing additional function calls.
|
This established workflow was successfully transitioned to NexusGenAI, culminating in a user interface that encapsulates these functionalities.
|
Okay, I’ll be the first to say it — the UI still needs some polish. When I was putting together the MVP, my main priority was nailing the functionality and making sure the user experience was top-notch. Aesthetics took a bit of a backseat. Nonetheless, the app does a perfect job of mirroring what NexusTrade’s chat does, minus the post-processing bits that are still hanging out in NexusTrade. The cool part? This separate setup means I can now tweak and add new prompts with ease — no more digging into the code for every little change.
|
For handling the prompts themselves, I came up with this brand-new UI:
|
For anyone who’s used the old OpenAI Playground, you’ll notice some similarities in this UI. It’s designed to be user-friendly, allowing users to give specific instructions tailored to their application’s needs. Plus, there’s flexibility in choosing the model: whether it’s GPT-3, GPT-4, or any of the other robust open-source options available. Users have the freedom to select what works best for them.
|
There are several other key enhancements too. Take, for instance, the ability to include a schema for a JSON object within a prompt, like this:
|
Based on the user’s input and the ensuing dialogue, the model is capable of producing a JSON that aligns with the specified schema.
|
Additionally, there’s an option for users who need their responses in a specific format: they can set the model to consistently provide valid JSON outputs.
|
Finally, users now have the ability to save examples to a prompt. This is a crucial feature that helps in steering the model towards delivering more relevant and accurate information.
|
Enabling More Complex AI Workflows
|
Considering NexusTrade has over 1,150 users, it’s clear that maintaining feature parity is very important. However, the real value of NexusGenAI lies in its ability to incorporate more sophisticated functionalities, like an integrated Retrieval Augmented Generation (RAG) Pipeline.
|
Retrieval Augmented Generation revolutionizes how we handle user queries by dynamically sourcing additional context. This process involves three steps:
|
- Uploading relevant documents to the application.
- “Chunking” those documents and saving their vector representation to a database.
- Fetching relevant chunks in response to each user request.
|
NexusGenAI seamlessly integrates a RAG pipeline, empowering users to refine model responses with context-specific information. For instance, by uploading my codebase, I can retrieve pertinent code snippets while seeking AI assistance with programming issues.
|
This integrated feature significantly amplifies the app’s capabilities, allowing for more nuanced and contextually aware responses.
|
Preparing for Broader Deployment
|
Even with just a few months of development, NexusGenAI has become a very feature-rich platform. Yet, there are a few important milestones that I must accomplish before releasing to a broader audience.
|
Enhanced UI
A professional appearance is key for widespread adoption. The UI needs a refined aesthetic with cohesive brand colors and themes, especially to appeal to large businesses seeking a polished product.
|
API Key Usage Monitoring
NexusGenAI is designed to be used via an API as well as the easy-to-use user interface. In order to accurately calculate costs, it is essential to implement a system to track API key usage.
|
Proving the Value with Another Use Case
It’s amazing that the NexusTrade chat is able to be configured within the platform. But, NexusGenAI isn’t just for AI-Powered Finance applications; it’s supposed to be useful for a wide array of different applications. Thus, my goal is to prove this to be the case, and develop another AI application using NexusGenAI for the backbone.
|
The AI and LLM landscape is rapidly evolving, necessitating user-friendly tools for managing AI applications. NexusGenAI is a stride forward, simplifying the configuration of complex workflows through an intuitive interface, proving indispensable for both developers and end-users alike.
|
Thank you for reading! If you’re interested in using AI to automate your existing processes or develop a customized chat with a wide range of models, sign up for NexusGenAI today. If you’re interested in how AI can be applied to finance, check out NexusTrade!
|
|
