Google’s Code Assist Enterprise can be trained on internal source code, is focused on Google services

Google’s Code Assist Enterprise can be trained on internal source code, is focused on Google services

Google has introduced Gemini Code Assist Enterprise, an AI coding assistant which can be trained on internal polices and source code. The product was announced at the Google Cloud Summit in London this week.

Code Assist in a coding assistant which uses the Google Gemini AI model to power its intelligence. The big feature of Code Assist Enterprise is that code suggestions can be based on internal source code repositories and policies. Code customization connects to either GitHub or GitLab repositories, with support for additional providers including self-hosted and on-premises promised for “early 2025”.

There are options to exclude files that should not be indexed, such as sensitive data, old or deprecated code, or auto-generated or temporary code. One assumes that with “auto-generated” Google has in mind code automatically generated by designers and wizards, rather than by Code Assist itself. It is also possible to restrict specified developers to a subset of repositories va IAM (Identity and Access Management) policies.

According to Google, Code Assist Enterprise does not train its Gemini model using organizational data, and code customization data lives in a Google Cloud managed project isolated to each customer’s organization.

Other new features unique to Code Assist Enterprise are AI assistant for Google Apigee (API management), AI for automation flows in Google Application Integration, and AI for Google BigQuery. Generic SQL support and AI for Google Firebase is in both Standard and Enterprise. The company has  a table of features showing the differences.

Code Assist Enterprise is available now and costs $19 per user/month on an annual subscription, until March 31, 2025, when the price will rise to $45 per month per user. Code Assist Standard is free until November 8th 2024, after which it also costs $19 per user/month for an annual commitment, with no free tier.

The market for AI coding assistants is a crowded one, and to date Google’s offering has received a mixed reception. The Visual Studio Code add-in for Code Assist, with is combined with other tools for GCP (Google Cloud Platform) as Cloud Code, has has over 1,400,000 installs but only three stars, with some developers complaining that “it just isn’t ready for prime time” and that “it is a work in progress. Should not be live in current state,” though others state that it works well with Google Cloud Shell and that it is an “excellent VS Code add on to work with Google Cloud.”

The company does seem to have a focus on its own platform and it seems likely that those who are already GCP (Google Cloud Platform) customers are the main market for Code Assist.

A brief look through the developer reviews for various coding assistants for VS Code, the most popular programmer’s editor, shows notable differences in ratings and also suggests that this product category is not yet mature.  Codeium, for example, has five stars and generally positive reviews, with a similar install count to Code Assist; whereas GitHub Copilot has over 20 million installs but mixed reviews, especially for recent updates. 

Despite the hype around AI coding help, it seems that these products are not yet mature and it is worth trying several to find out which one works best in the context of a particular developer’s work.