Continue.dev、VS Code拡張機能 v1.2.23-vscode をリリースContinue.dev Releases v1.2.23-vscode

Continue.dev has published version 1.2.23-vscode, a stable update to its open-source AI coding assistant for Visual Studio Code. The release matters because Continue is one of the more widely used open-source alternatives to proprietary pair-programming tools such as GitHub Copilot, and changes to its onboarding and default configuration directly affect how new users first connect the extension to a language model.

According to the release notes, the update removes the CLI-install banner that previously surfaced inside the extension, drops the Generate Rule feature, and switches both the onboarding flow and the new-configuration template to an explicit model-definition approach. Taken together, the changes appear aimed at simplifying the first-run experience while making the way models are declared more transparent.

For readers unfamiliar with the project, Continue is an extension for VS Code and JetBrains IDEs that adds chat, inline editing, and autocomplete backed by large language models. Unlike closed assistants tied to a single backend, Continue lets developers choose their provider, whether a commercial API such as OpenAI, Anthropic, or Mistral, or a locally hosted model served through tools like Ollama or LM Studio. Behaviour is driven by a configuration file, historically config.json and now typically a config.yaml, in which users declare the models they want and the roles those models fill, such as chat, edit, autocomplete, and embeddings.

The move to explicit model definition is the most consequential change for everyday use. Rather than leaning on implicit defaults or an automatically selected model, the onboarding process and the starter template now prompt users to spell out the specific model, provider, and role they intend to use. The likely benefit is reproducibility: a configuration that names its models outright is easier to share, debug, and reason about, and it reduces ambiguity about which model is actually answering a given request. The trade-off is a modest increase in upfront effort, since first-time users must make an explicit choice rather than accepting a preset. This direction is consistent with Continue's broader emphasis on configuration-as-code and user control over which models run where.

The removal of the CLI-install banner trims a piece of promotional interface that pointed users toward Continue's separate command-line tool. Continue maintains a CLI for working with its assistant outside the editor, and removing the in-extension prompt reduces visual clutter without necessarily signaling any change to the CLI itself. The retirement of Generate Rule, a feature that helped users create reusable instruction "rules" for the assistant, is harder to interpret from the notes alone. Removing a feature in a stable release often reflects low usage, maintenance burden, or an intention to redesign the capability later, though the release does not state a reason, so any explanation remains speculative.

Rules themselves are worth understanding as background, because they are not unique to Continue. A rule is a persistent instruction that shapes how the assistant responds, for example enforcing a coding style, a preferred framework, or project-specific conventions. Comparable mechanisms exist elsewhere in the ecosystem, including custom instructions in GitHub Copilot and project rules in editors such as Cursor. Removing only the automated generation of rules does not necessarily mean the underlying rules system is gone; it appears to target the helper that produced them rather than the concept as a whole.

The release lands amid steady iteration across the AI developer-tooling space, where vendors are competing on model choice, context handling, and the friction of setup. Continue's positioning has long rested on openness and flexibility, letting teams keep code and inference local when privacy or cost demands it, and an explicit configuration model fits that stance.

As with any incremental update, the practical advice for existing users is to review the changelog before upgrading, particularly if their workflow relied on Generate Rule or on implicit model selection during setup. Users who pin specific versions or maintain shared configurations across a team should confirm that their existing config files still validate against the updated template. The full list of changes is available through Continue's release channel on its repository.