portal-ai-plugin

Problem

Asking a model about the code in front of you usually means leaving the editor, losing the context, and pasting fragments into a browser tab.

Approach

A JetBrains plugin in Kotlin on JVM 21, backed by a local Ollama instance rather than a hosted API, so the code never leaves the machine. Conversations are split into named channels, which keeps a design discussion from being buried under a week of dev notes.

Outcome

• Distributed through the JetBrains Marketplace, installable from the IDE plugin browser.
• Runs entirely against a local model: no key, no account, nothing sent off the machine.
• The endpoint and model are configurable, so it points at a shared Ollama server as easily as at localhost.