Feature Overview

Surface private generative AI features (similar in function to ChatGPT) via a plugin in Red Hat Developer Hub.

Refinement Document

Goals (expected outcomes)

Make it easy for any user of Red Hat Developer Hub (in any role) to get assistance in their work from AI using company sanctioned, private large language models.

This assistance could be very flexible, potentially helping and assisting with a very wide variety of tasks (including, but not limited to, technical tasks). Currently, because of privacy and security concerns, many organizations have restrictive AI policies that limit the ability of development teams to leverage AI.

By creating a safe AI solution, we can help teams overcome these objections and unlock additional productivity benefits. By surfacing this feature in RHDH, we can help the product grow and also boost retention.

Requirements (aka. Acceptance Criteria):

A list of specific needs or objectives that a feature must deliver in order
to be considered complete. If the feature spans across releases then good
to have scope for each release with acceptance criteria. Be sure to
include nonfunctional requirements such as security, reliability,
performance, maintainability, scalability, usability, etc.

• The plugin should be clearly owned by red Hat and branded where appropriate to help enforce that relationship (for example with the plugin name such as "Red Hat Developer Hub AI Plugin" or with graphical branding of some kind)
• No guarantees should be made as to the security or performance of the plugin, unless possible and necessary due to the implementation choices we make.
• The most appropriate (loosest) release status should be chosen (e.g. "Developer Preview" to make it clear that this feature is early and limited in scope. 

Out of Scope (Optional)

• It is not necessary for the feature to have it's own UI if existing UI's can be surfaced in RHDH or reused from elsewhere.
• It is not necessary for the feature to serve, bundle, or embed it's own LLM - we can use OpenShift "Single Model Server" or similar for that.
• It is not necessary for the feature to provide or depend on a formal model registry. Models servers tend to have solutions for that already.

Customer Considerations (Optional)

Customers will need to provide (install, expose, etc) a private AI chat server (an inferencing server).

Documentation Considerations

• Documentation will be required to show how to configure this feature.
• Documentation will be required to help customers to provide at least one private generative AI chat server using either OpenShift AI or existing open source (e.g. Ollama, vLLM, etc.)