Getting Started

xpreiIDE is a bring-your-own-model AI coding assistant. Chat with a local Ollama daemon or any OpenAI-compatible endpoint (OpenAI, vLLM, LM Studio, OpenRouter, Together, …) — no vendor lock-in, no forced cloud.

1. Start a model backend

The easiest option is a local, free one:

ollama serve
ollama pull llama3.1

Any OpenAI-compatible endpoint works too — you’ll need its base URL and an API key.

2. Install the extension

Grab it from the VS Code Marketplace or build from source, then open the xpreiIDE panel (it also opens automatically the first time you install it).

3. Pick a model

You usually don’t need to do anything else. If a local Ollama daemon is already running when the extension starts, it’s detected automatically — several installed models triggers a one-click setup notification. If nothing was detected, run xpreiIDE: Select Model from the Command Palette (Ctrl/Cmd+Shift+P) and pick a provider and model.

4. Chat

Type a message and press Enter. The reply streams in token by token. From here, see Chat & Modes for Plan/Edit/ Agent modes, Context Mentions for pulling in @codebase/@file/etc. context, and The Agent for the full autonomous multi-file loop.

Adding a hosted / custom model

Use xpreiIDE: Add Model Provider for guided setup — common presets plus a “Custom…” option for anything speaking the OpenAI-compatible chat API. You’ll be asked for a base URL (include the API version segment, e.g. /v1), an API key if needed (stored in your OS’s secure credential store — Windows Credential Manager, macOS Keychain, etc. — never written to a plain-text file), and optionally a default model name.

Advanced: providers (minus API keys) live in one shared, human-editable file at ~/.xpreiide/config.yaml — the same file the JetBrains and Eclipse plugins read, if you use those too:

providers:
  - id: ollama-local
    kind: ollama
    label: Ollama (local)
    baseUrl: http://localhost:11434
  - id: openai
    kind: openai-compat
    label: OpenAI
    baseUrl: https://api.openai.com/v1

Per-role models

By default, chat, ghost-text completions, the agent, inline edit, and commit-message generation all use whatever model xpreiIDE: Select Model set for chat. Run xpreiIDE: Select Model for Role… to use a different model for one of them — for example, a small, fast local model for completions while a larger model drives the agent. Roles you haven’t overridden keep following the chat model automatically.