Introduction

Bel Esprit is your personal AI solution architect, designed to simplify the process of building, customizing, and deploying AI solutions. This user guide will walk you through the key functionalities and interactions with Bel Esprit.

Note: Bel Esprit is currently in its alpha version. To gain early access, you can sign up for the waitlist and be among the first to experience it. Once granted access, simply log in to the aiXplain platform to start using Bel Esprit.

Build Your First AI Solution 

Bel Esprit is a Generative AI chatbot designed to understand natural language instructions and build deployable AI solutions based on them. 

To start of, here are some examples:

• "Build me a video summarizer solution."

• "I want a solution to dub my French podcasts to Spanish."

• "I need a system to analyze the sentiment, topic, and keywords in my product reviews."

Bel Esprit analyzes your instructions, selects the appropriate AI models for your solution, then it builds and validates the architecture of your solution before displaying it onto aiXplain’s Design canvas. If a requirement is missing from your instructions, Bel Esprit would ask you follow-up questions until it has the necessary information to build your solution.

Example

User instruction: I want to add a sales chatbot that can also detect sentiment on my US e-commerce website

Generated AI solution: 

Modify Your Solution

You can ask Bel Esprit to add or remove a functionality, change the used models, or support other languages and data modalities. Additionally, the user has the option to manually modify the solution within aiXplain's Design canvas. It's important to note that in the current version of Bel Esprit, manual modifications of the solution will disengage Bel Esprit from further involvement, and it will not be able to provide follow-up on the modified solution.

Example

User instruction: I also want to detect any hashtags in the input

Re-generated AI solution:

If your solution contains Text Generation models, Bel Esprit attempts to extract the appropriate ‘prompt’ to tailor these models for your use case. If this information is not provided or is unclear in your instructions, Bel Esprit will follow-up with you with questions. 

In the example above, the shales chatbot, Bel Esprit was able to detect the ‘prompt’ for the Text Generation model from the user’s instruction and set it to ‘Act as a sales chatbot for a US e-commerce website’.

Change AI Models

You may want to explore other AI models before committing to a solution. Bel Esprit can search and filter AI models in aixplain’s marketplace.

Ask Bel Esprit:

• “List all text generation models available on aiXplain”

• “Show me english to arabic translation models from Azure”

• “Replace the current Speech Recognition model in my solution by another one”

Example

User instruction: Use OpenAI text generation model in my solution

Re-generated AI solution:

Test Your Solution

If your solution is complete, you can ask Bel Esprit to test it and try it out. Bel Esprit will save your solution under ‘Pipelines’ on your Dashboard. It will then take you to the ‘Try it out’ page in aiXplain’s platform. Here, you can input or upload sample data to see how your solution behaves. 

Ask Bel Esprit:

• “Let’s test this”

Deploy Your Solution 

When a solution is saved, a deployment endpoint is immediately generated and your solution is automatically hosted in aiXplain’s platform.  

If you are ready to integrate your solution into your own code, application or notebook, you can ask Bel Esprit for the integration code. The code will include your deployment endpoint. Copy that code into your codebase to deploy and use your solution from outside aiXplain’s platform.

Ask Bel Esprit:

• “What are the next steps to deploy my solution?”

Handle Incomplete solutions

If Bel Esprit is unable to build your solution or parts of it, it will recommend a few options. One option is to use a Generative AI model in place of some missing models. While these models won't produce the highest quality output and may be more costly, they allow for quick testing of your solution.

If this step does not result in a complete, deployable solution, Bel Esprit will offer to connect with a certified aixplain Specialist to help you finalize your solution.

Contact A Specialist 

We have made the process of contacting specialists frictionless via Bel Esprit. No matter the type or size of your AI challenge, our certified AI specialists can help you tackle it in a short time. 

Ask Bel Esprit:

• “I need a specialist to help make my solution product-ready”

• “I want to stress-test my solution, connect me with a specialist”

When you request to contact a specialist, Bel Esprit generates the necessary Request for Proposals (RFPs) to build the missing AI components based on your conversion. It then sends these RFPs to our network of specialists. Once Bel Esprit finds a match, Bel Esprit contacts you via email to facilitate this communication.