FAQ

You will find the most common questions and answers about StackSpot AI.

StackSpot AI

What are the main differences between StackSpot AI and Competitors?

Context-Based Code Generation: StackSpot AI focuses on hyper-contextualized code generation. The AI considers the request's text and the context of the project and the company, resulting in more precise and developer-aligned code.
Knowledge Sources: We have a unique feature is the ability to create and share "Knowledge Sources." This allows companies to capture and use domain-specific knowledge to enhance their code generation capabilities.
Stack AI: We ensure that code generation aligns with the company's technological vision. StackSpot AI supports various programming languages and technologies, making it versatile and practical for developers in different technological stacks.
Custom Quick Commands: This feature streamlines interactions with the LLM, enabling users to execute predefined commands quickly and efficiently, thereby improving productivity during the coding process.
Enterprise Features: StackSpot AI was developed with a focus on corporate needs. It offers businesses access control, permission management, and security features.
Greater Customization and Control: StackSpot AI enables companies to have greater control over their AI, which is crucial when compliance, security, and specific company requirements are paramount. Users can tailor their AI to meet the particular needs of their project or company, which is particularly beneficial in corporate environments where coding requirements can differ significantly.
Influence on Development Roadmap: StackSpot AI has a continuous development Roadmap, meaning the platform will continue to evolve and improve more efficiently than other big tech companies based on market needs and user/customer feedback.
Integration with StackSpot EDP: They are designed to seamlessly integrate with the StackSpot Enterprise Developer Platform (EDP), allowing companies to incorporate EDP contents to generate contextualized code in StackSpot AI.
SCMs: There is no requirement for the project to be hosted on a specific SCM for StackSpot AI to function fully

Where is the data stored by StackSpot AI, such as Knowledge Sources and Stack AI? How do we ensure the security of this information?

We have two account models with different deployment approaches: the Shared Model and the Private Model. Depending on the model, it works differently.

Shared Model: Each account has its isolated database schema (Vectorized DBs, Relational DBs, and Non-Relational DBs). All shared account models use the LLM instances provided by StackSpot AI.
Private Mode: Each account has its private instance of the database (Vectorized DBs, Relational DBs, and Non-Relational DBs) All private account models will have their private instance of the LLM from StackSpot AI.

What are the confidentiality precautions for the codes we will use to provide context to the application?

The StackSpot AI LLM is private within companies, ensuring that the content shared in conversations with the agent will not be exposed. Confidential codes will not be made global.

Do you use information from the open project in the IDE to generate context for the responses brought by the LLM?

We utilize user project data to improve the LLM's capability to generate code with personalized context. By examining and understanding specific project details such as programming language, used libraries, and technical requirements, we can offer more efficient solutions tailored to individual needs. It creates more accurate, effective, context-aligned codes, ultimately enhancing productivity and software development quality. Users can select two indexing strategies based on their requirements:

1. Selective Indexing: Users can index only the project's most relevant files, as determined by the StackSpot model. This approach provides a simple yet powerful and functional indexing method. 2. Full Project Indexing: Users can index the entire project, including all its files. This option allows for comprehensive indexing of the project's content.

Do OpenAI's assistants (RAG) perform tasks similar to those of StackSpot AI's knowledge sources? What are our differentiators?

OpenAI's assistants and StackSpot AI's knowledge sources offer relevant and contextualized information but with different approaches and capabilities. StackSpot AI has simplified integrating Knowledge Sources, making it more efficient and accessible through an intuitive portal. Users can quickly load Knowledge Sources using methods like copy-pasting and file uploads to incorporate various Knowledge Sources in different workspace contexts, leading to satisfactory results. Knowledge Sources in StackSpot AI are divided into 'chunks' to optimize content utilization, enabling refined and relevant searches and ensuring accurate and contextualized responses from the Language Learning Model (LLM). StackSpot AI focuses on prompt engineering, continually refining prompt construction methods to enhance operational effectiveness and reflect a commitment to improving the user experience. The system ensures secure and effective information management and distribution to different teams, resulting in personalized and relevant responses from an agent configured exclusively for each team. In summary, while OpenAI's assistants offer advanced AI-based search and content generation, StackSpot AI provides tailored technologies for enterprise use, including personalized knowledge management, access control, prompt optimization, and easy integration of Knowledge Sources into diverse work environments.

How does it work with modernization cases using StackSpot AI?

You can leverage our Knowledge Source feature to input legacy code. Combined with Stack AI, where you define the technological vision of the outputs generated by AI, you can make inquiries based on legacy code and generate modernized code based on your TechView.

Is the StackSpot generative AI assistant native, or do you use a market API?

We use OpenAI's LLM. StackSpot AI leverages whatever you need when it comes to LLM models. OpenAI is the model we offer in our solutions, and we also deliver LLM (GPT 3.5 Turbo and GPT-4), depending on the instance you are using. Soon, we will use GPT-4 Turbo in some features that coexist with GPT-4.

Is there a StackSpot AI plugin for IntelliJ?

Yes, install it via the IntelliJ's Software center.

What are the advantages of StackSpot AI compared to competitors?

StackSpot AI offers unique features, including Stack AI, Quick Commands, and Knowledge Sources. These features speed up your prompts and provide context to the AI by incorporating specific company information using advanced RAG and chunking techniques. StackSpot AI also utilizes the context of your project.

Why Should I use StackSpot AI instead of competitors?

StackSpot AI is a specialized code assistant focused on technology development. It provides high-quality responses and allows the addition of contextual layers, such as Stacks AI and Knowledge Sources, to understand the specifics of your work. This level of contextualization ensures accurate answers, even without advanced prompt engineering techniques.

Another unique feature of StackSpot AI is the Quick Commands, which automate tasks with chained prompts and web requests. You can save prompt sequences in StackSpot for future use, powered by Knowledge Sources and web requests. Additionally, integrated Quick Commands are offered for standard technology tasks, such as creating automated tests, documenting code, explaining code, identifying vulnerabilities, and refactoring code snippets.

Can StackSpot AI be used in Visual Studio Enterprise 2022?

The extension is not yet compatible with Visual Studio Enterprise, but we are working to enable it. We do not currently have a defined delivery date, but you can follow the progress on our Public Roadmap: Visual Studio Extension for StackSpot AI.

Does StackSpot AI perform equally well in Portuguese, or is it better in English?

The product is designed to work equally well in Portuguese; however, due to best practices in prompt engineering, some users report better results when using prompts in English or when rephrasing the prompts.

Can StackSpot AI also assist platform teams that develop GitHub Actions Workflows and deal with IaC tools (CloudFormation and Terraform)?

Yes, you can use Stacks AI, Quick Commands, and Knowledge Sources with specific context for this purpose.

Is there any feature mapped for StackSpot AI to read images?

This feature is currently in the Discovery process. Feedback on its use cases and applicability in clients' daily activities would help us reassess and prioritize it more effectively within our roadmap.

Can I use StackSpot AI as a manager in my daily activities?

You can use StackSpot AI to generate user stories, refine issues, and conduct PDI evaluations. Soon, we will launch the functionality to create specialized agents, further enhancing these activities.

To ensure better results, we suggest adding examples to our Knowledge Sources and using Quick Commands for prompt engineering. We will soon introduce the ability to create specialized agents, further enhancing this process.

I used StackSpot AI to improve the quality of user stories. This needs to be done via VSCode or another IDE. Are there plans to provide a web interface to facilitate business use?

Currently, we have a "Try It" section for interaction with the StackSpot AI Assistant via the Portal. However, the experience via IDE is more comprehensive. We are evaluating the development of a more robust solution in the StackSpot AI Portal.

Does StackSpot AI support XCode (iOS)?

After assessing the situation, we have found compatibility limitations that make developing an extension for Xcode more complex. Moreover, we need to guarantee continuous support and evolution of the extension. As a result, we have yet to begin development, but we plan to start soon.

Can I use the StackSpot AI plugin in Visual Studio 2022?

Currently, the extension is not compatible with Visual Studio Enterprise, but we are working to enable it. We do not have a defined delivery date at this time, but you can follow the progress on our Public Roadmap.

Is it possible to adjust the LLM temperature?

At present, this configuration is not currently available. We have a backlog issue related to temperature adjustment. Please inform us through our feedback channel if you find this necessary.

Which LLM StackSpot AI uses?

It uses the most recent GPT models - OpenAI. But, it will have its instance soon.

Do you have a support channel?

Yes! Send your questions to StackSpot AI support email.

StackSpot AI Concepts

Why use the concept of Workspaces?

Developers can take advantage of pre-configured environments and access specific resources designed to accelerate the development of applications for a targeted group of people in a particular Workspace. The result is similar to having a customized and focused agent that caters to the determined audience in that Workspace.

Do I have to use a Workspace?

No, but you should. Creating or choosing a Workspace will help StackSpot AI get more context and share better answers.

How does Stack AI work?

It works as "Declarative Stacks," where the governance team declares the technologies to be used in the project within a Workspace via a portal form. These should be derived from patterns found in actual code, ensuring that the user aligns with the technological vision of CTOs/Architects to ensure the implementation of best practices from the company's perspective and enhance developer productivity.

What is a Stack AI?

It's a group of technologies you can create, and it is used to give more context to LLM and generate better codes. It contextualizes the user input with specific user definitions.

How can I open a new chat prompt on StackSpot AI?

Access your IDE:

Visual Code Studio

On the upper menu, click on the '…' icon;
You will see an ‘Options’ tab. Click on ' New chat'.

Knowledge Sources

What does Knowledge Source do?

When you add Knowledge Source, you feed the platform's context. It is not AI training; it works with the information you insert. It searches for similarities in the database to give you better codes.

How long does it take for a Knowledge Source to work after I created it?

It doesn't take too long. After processing, you can ask for help on the StackSpot AI extension.

I sent files to my Knowledge Sources. Can I delete them?

Yes, to delete them, access your Knowledge Source, click on the 'Manage Knowledge Source' button, and choose which one you want to delete. You can also delete Knowledge Objects individually by clicking on the delete button for each object.

File uploads to the Knowledge Sources got stuck. What should I do?

Wait until the upload finishes. If the error persists, click on 'View details' to access what has been processed and check if any files have encountered errors. In that case, correct the incorrect file and upload it again.

Do I need to select the Knowledge Sources every time I use StackSpot AI?

Yes, you need to select all desired Knowledge Sources after logging in to StackSpot AI.

Can I edit a Knowledge Source?

Yes, Knowledge Sources within Personal can be edited. However, KSs (Knowledge Sources) in the Account can only be edited by the account admin. Editing includes the ability to delete, rename, and change the description of your KS.

Can I edit Knowledge Source Objects?

No, if you need to add more files to your Knowledge Source Objects, delete the existing ones and upload the file again.

Is my Knowledge Source’s information visible to other StackSpot clients?

No. Only in your account.

How do Knowledge Sources work?

We use vectorized databases with two types of content: APIs and client Code Snippets. The intention is to enhance the context generated for the LLM, increasing developers' efficiency as they use information sources referenced by the organization.

Details of API Knowledge Sources:

Clients can bring their API catalogs into the platform to be used as context for the platform to provide integrations with existing APIs. To add Knowledge Objects of the API type, OpenAPIs (swaggers) must be imported into the StackSpot AI platform.

Details of Code Snippet Knowledge Sources:

Clients can add and use custom code snippets in StackSpot AI, which provide specific functionalities, demonstrate coding styles, and enrich the language model's learning process with real-world scenarios. This functionality allows for personalized coding experiences and encourages collaboration within the platform. To add Knowledge Objects of the Code Snippet type, users must bring the Code Snippet and its "Use Case" into the platform.

Can we integrate Knowledge Sources through an API?

You can use dedicated endpoints for batch and individual Knowledge Sources uploads.

Are there limits on Knowledge Sources per user? Are there size limits for KS?

There are no limits on the number of Knowledge Sources (KS) per user for Enterprise accounts.

Whenever I change an API contract or create a new API, do I have to update the Knowledge Sources manually?

Currently, yes. However, we are creating connectors that allow independent updates (fetch/update) of Knowledge Sources from specific sources with codes or OpenAPIs.

Can the tool identify opportunities for reuse and application of classes or modules already used by the team through Knowledge Bases?

Knowledge Sources do not suggest reuse opportunities. We recommend using Project Indexing and prompt generation for project improvements. Code suggestions, another feature on our roadmap that may help, are also on our roadmap.

Can incode be registered as a Knowledge Source?

We do not currently have Knowledge Sources for Incode, but we are already using StackSpot AI with COBOL in some initiatives.

*How can I provide API examples for my KS? Should I include the complete API files or just code snippets? What is the best approach?**

Use API-type Knowledge Sources by providing OpenAPI contracts.

Is it possible to use RAG with the plugin to query a document in context?

You can make the document available via Knowledge Sources and add it to the context.

Quick Commands

What are Quick Commands? How do Custom Quick Commands work?

Quick Commands are pre-defined prompts that allow you to send commands to the LLM quickly via the right-click menu. This feature makes it faster and easier to interact with the LLM, letting users execute commands swiftly and efficiently, which improves productivity during the coding process. StackSpot AI users can use Quick Commands provided by StackSpot AI or create their own Custom Quick Commands (CQC). To create CQC, users can develop workflows using input and output prompts and inputs and outputs from Requests/Fetch HTTP to create a flow that can be executed quickly and effectively by the Custom Quick Command. It helps to boost productivity and ensure a great developer experience.

Using Quick Commands, can a flow with two distinct outputs be created?

We currently don't have these features. We're planning design sessions to improve the Quick Commands flow, including multi-selection of inputs at the start of execution, conditionals, and interactive inputs. You can follow our Public Roadmap to stay updated on the latest developments in Quick Commands.

Is it possible to create a Quick Command to perform integrated tests? For this, would it be necessary to analyze the entire project, not just a snippet of code?

You can ask for assistance from StackSpot AI by using prompts to generate comprehensive tests. To make this process easier, we suggest creating Quick Commands.

Permissions

Can anyone create a Workspace?

Currently, only administrators can create content. We ask users to create only personal content until the Content Management Plan for StackSpot AI in companies is released.

How is accessibility data captured for use in a Dashboard? Do all users who use the Quick Command need to send the data somewhere, or is it done automatically?

To use the Quick Command, select the code you want to evaluate, right-click, and navigate to "StackSpot AI -> My quick commands -> your quick command name."

How is the Code Reviewer being used in native development?

The Code Reviewer is not a standard Quick Command in StackSpot AI. However, users have achieved excellent results by requesting Code Reviews via the prompt. Based on these results, specific Quick Commands for Code Review have been created and are available in the "Account" tab for use in the IDE. Due to its success and demand, this functionality will become an integrated feature in the platform.