Crafting Intelligent Workflows: A Guide to AI Powered Assistants

Securing and speeding up your back-end development

Securing and speeding up your back-end development

Not since the invention of the World Wide Web itself has humanity seen such a dramatic cultural shift as we're currently living through with the widespread availability and adoption of AI. This technology will (and has) forever changed the way we work, live, and exist in this world. While it can be difficult to keep up, it's important to appreciate the significant evolutions in the AI space as it builds momentum, and one of those is the potential of AI Agents.

Throughout the video series, I showcase the practical steps of how BuildShip's platform can be utilized to integrate OpenAI's Assistant API. The process involves intuitive, visual tools that abstract away the complexities typically associated with such integrations. Users can visually map out their workflows, connect to various data sources, and embed AI-driven interactions without delving into the underlying code. This not only simplifies the development process but also allows for rapid prototyping and iteration, which is crucial in the fast-paced world of technology.

The Beginning... OpenAI's Chat GPT

OpenAI has been at the forefront of AI research and deployment, significantly advancing the field with its development of Chat GPT. Built on the Generative Pre-trained Transformer models, Chat GPT has laid a robust foundation for creating AI assistants capable of understanding and generating human-like text. This breakthrough technology has not only made AI more accessible and customizable but has also opened new avenues for developing AI assistants that can perform a wide range of tasks, from answering queries to providing recommendations and executing commands. However, Chat GPT was just the beginning of a broader exploration into the potential of AI assistants.

Chat GPT served as many people's first glimpse into the capabilities of conversational AI, showcasing the potential for these technologies to transform everyday tasks and interactions. It demonstrated that AI could go beyond simple command-response mechanisms to engage in more nuanced and contextually aware conversations. This initial exposure has sparked widespread interest and investment in the AI assistant space, leading to rapid developments and innovations.

Today, we are witnessing the emergence of AI assistants that are not only more sophisticated but also more specialized and integrated into various aspects of work and productivity. These advancements are driven by the need for more efficient, personalized, and context-aware interactions with technology. AI assistants are now being designed to understand user preferences, learn from interactions, and even anticipate needs, making them more than just tools for information retrieval or task execution. They are becoming collaborative partners that enhance human capabilities and streamline workflows.The evolution from Chat GPT to today's AI assistants represents a significant leap forward in AI technology. It reflects a shift from a focus on the technical feasibility of generating human-like text to the practical application of AI in enhancing human productivity and creativity. As AI assistants continue to evolve, they are set to revolutionize the way we work, learn, and interact with the digital world. This rapid development in the assistant space underscores the dynamic nature of AI technology and its potential to reshape our relationship with machines, making it an exciting area of ongoing research and innovation.

 

The Evolution: Leveraging BuildShip's Low-Code Platform

BuildShip represents a significant advancement in the realm of application development, particularly in the integration of AI capabilities. As a low-code platform, BuildShip is engineered to empower a wide range of users—from seasoned developers to business professionals with limited coding expertise—to construct sophisticated app backends, automate complex workflows, and develop APIs with minimal coding effort. This approach not only accelerates the development process but also lowers the barrier to entry for leveraging advanced technologies in custom applications.

Getting Started: Building a Basic AI Assistant

1. Set Up with OpenAI

Begin by visiting https://platform.openai.com, sign in, and navigate to the 'Assistants' section to create a new assistant. Give it a name, craft a prompt, select a model, and take note of the assistant ID.

2. Configure in BuildShip:

In BuildShip, go to 'Templates' and choose the pre-built workflow for the AI assistant you've just created in OpenAI. Input your OpenAI key (stored securely in Secrets) and the assistant ID from OpenAI.

3. Detail Your Instructions:

The instructions you provide in BuildShip are crucial as they will override the default instructions in OpenAI. Define the assistant's role, the tone and manner of response, and any additional resources it should use.

4. Engage with Your Assistant:

Once you've hit 'Ship', you can start interacting with your assistant. To continue a conversation, use the thread ID provided to maintain context.

5. Expand and Refine:

As you become more familiar with the AI Assistant Builder, you can refine your assistant's capabilities by adding context from documents and integrating tools you already use within BuildShip.

Real-World Example: A Personal Tutor

Take a look at the example given on the BuildShip website: https://docs.buildship.com/ai-models/assistant#use-case-i-chat-with-assistant

Imagine you've built a personal tutor assistant. You ask it to explain the Northern Lights concisely, and it delivers. To further the conversation, simply include the thread ID in your next query. Ask where you can watch the Northern Lights from, and the assistant, already aware of the context, provides an accurate response.

Level Up: Document Retrieval and Analysis

Document Retrieval allows you to tap not only into OpenAI's vast knowledge base, but also reference and learn from your own documents.

1. Selecting the Template

Navigate to the **template section** within BuildShip and choose the **Assistant with Retrieval template**. This template is your ticket to creating a chat assistant that can provide context-aware responses based on the files you upload.

2. Understanding the Limitations

Before you proceed, note that there are some limitations to keep in mind:

  • Files cannot exceed 512 megabytes
  • Document size should be around **2 million tokens

3. Setting Up Your Assistant

  • Upload Contextual Files: First, upload the files that will serve as context for your assistant into OpenAI. For instance, a summary report titled "Augmenting Human Intellect: A Conceptual Framework" can be used.
  • Create a New Assistant: Go to `platform.openai.com` and create a new assistant. Give it a friendly name for easy identification.
  • Choose the Model: Select ChatGPT 4 or Turbo preview as the model and enable the retrieval feature.
  • Upload the File: Now, upload the file you want the assistant to reference.

4. Testing Your Assistant

Before integrating with BuildShip, test the assistant on OpenAI's platform to ensure it behaves as expected and references the information correctly.

5. Integrating with BuildShip

Copy the assistant ID from OpenAI and input it into BuildShip, making sure to turn on document retrieval in the built-in tools. Test it out by sending a message in JSON format and observe the assistant's response.

6. Continuing the Conversation

To pick up a conversation later, use the provided thread ID to maintain context. The BuildShip documentation offers a detailed example of this process.

Expanding the Use Cases

Imagine the possibilities with your customer system—menus, policy documents, periodicals, and more can now be part of the conversation. This method ensures that your assistant's responses are based on your data, not on external sources beyond your control.