Skip to main content

Frequently Asked Questions

General​

Is DevoxxGenie free?​

Yes, DevoxxGenie itself is completely free and open source. It uses a BYOK (Bring Your Own Keys) model — you supply your own API keys for cloud LLMs, or run local models with Ollama or LM Studio at no API cost. There is no DevoxxGenie subscription fee.

Does DevoxxGenie send my code to the cloud?​

Only to the LLM provider you explicitly configure:

  • Cloud providers (OpenAI, Anthropic, etc.): your prompts, which may include code, are sent to that provider's API under their own privacy policy.
  • Local providers (Ollama, LM Studio, etc.): nothing leaves your machine. The model runs locally.

DevoxxGenie itself does not collect, store, or transmit your code.

Which IntelliJ versions are supported?​

DevoxxGenie requires IntelliJ IDEA 2023.3.4 or later. It works with the Community and Ultimate editions, and with other JetBrains IDEs on the IntelliJ platform (PyCharm, GoLand, WebStorm, etc.).

Does DevoxxGenie work offline?​

Yes — if you use a local model provider like Ollama. Once you've downloaded a model, DevoxxGenie runs entirely offline. Cloud providers (OpenAI, Anthropic, Gemini, etc.) require an internet connection.

LLM Providers​

Which LLMs does DevoxxGenie support?​

Cloud providers: OpenAI (GPT-4o, o3, o4-mini), Anthropic (Claude 3.5/4), Google (Gemini 1.5/2.x), Grok (xAI), Mistral, Groq, DeepInfra, DeepSeek (R1, Coder), Kimi (Moonshot AI), GLM (Zhipu AI), OpenRouter, Azure OpenAI, Amazon Bedrock

Local providers: Ollama, LM Studio, GPT4All, Llama.cpp, Jan, Nativ (MLX on Apple Silicon), any OpenAI-compatible endpoint

How do I use Ollama with DevoxxGenie?​

See the full Ollama setup guide. The short version:

  1. Install Ollama and pull a model: ollama pull llama3.2 or ollama pull llama4
  2. In DevoxxGenie settings → LLM Providers → Ollama
  3. Leave the base URL as http://localhost:11434
  4. Click Refresh Models and select your model

Can I use my own API endpoint (OpenAI-compatible)?​

Yes. DevoxxGenie supports custom providers — any endpoint that speaks the OpenAI chat completions API, including self-hosted models, DeepSeek R1, Grok, JLama, and enterprise AI platforms.

Features​

What is Agent Mode?​

Agent Mode enables the LLM to autonomously explore and modify your codebase using built-in tools — reading files, listing directories, searching for patterns, running tests, and making targeted edits. Instead of manually providing code context, the agent investigates your project on-demand.

It works with both local Ollama models (e.g. Qwen2.5, Llama 4, Mistral Small) and cloud providers.

What is Spec-driven Development (SDD)?​

Spec-driven Development is a workflow where you define what needs to be built as structured task specs with acceptance criteria (stored as markdown files), and the LLM agent figures out how to build it.

The DevoxxGenie Specs tool window shows tasks in a Kanban board. Click "Implement with Agent" and the agent checks off acceptance criteria as it works.

What is MCP?​

MCP (Model Context Protocol) is an open standard that lets LLMs connect to external tools and services. DevoxxGenie includes a built-in MCP Marketplace where you can install servers for filesystem access, web browsing, databases, APIs, and more. The LLM uses these tools automatically during conversations.

What is RAG?​

RAG (Retrieval-Augmented Generation) indexes your project's source code into a local vector database (ChromaDB via Docker) using Ollama embeddings. When you ask a question, the most semantically relevant code snippets are retrieved and included in the prompt automatically — giving the LLM accurate project context without manual file selection.

What are Commands?​

Commands are reusable slash commands you define in settings. Type /explain, /test, /review, or any custom command in the prompt input to trigger a predefined prompt template. Built-in commands include /test, /explain, /review, /find (RAG search), and /init.

Note: this feature was previously called "Custom Prompts" and then briefly "Skills". It is now called Commands. The current Skills feature is something different — see below.

What are Skills?​

Skills are self-describing capabilities loaded from SKILL.md files on disk. They are activated by the LLM itself (not the user) via an activate_skill tool call when the model decides a skill is relevant to your request. Skills require Agent Mode and a tool-capable LLM, and DevoxxGenie scans six directories — including ~/.claude/skills/ and .agents/skills/ — so the same files work across Claude Code, .agents-aware tools, and DevoxxGenie.

What is inline code completion?​

Inline completion provides GitHub Copilot-style ghost-text suggestions as you type, powered by Fill-in-the-Middle (FIM) models via Ollama or LM Studio. Enable it in Settings → DevoxxGenie → Inline Completion.

Troubleshooting​

Ollama models aren't showing up in the dropdown​

Click Refresh Models in DevoxxGenie settings after pulling a new model. Make sure Ollama is running (ollama serve or verify http://localhost:11434 is reachable in a browser).

Responses are very slow with local models​

Switch to a smaller quantized model. For chat, try llama3.2:3b or llama4:scout. For inline completion, try qwen2.5-coder:0.5b. See the Ollama performance tips.

The chat UI doesn't load on Windows — how do I fix it?​

On some Windows systems with certain GPU/driver configurations, the DevoxxGenie chat UI may fail to initialize with an error about Skiko/Direct3D.

Good news: current DevoxxGenie versions force software rendering on Windows during startup, so the chat UI should avoid these GPU-specific failures entirely.

If you still hit the issue, restart IntelliJ IDEA to ensure the Windows startup override is applied.

If you are on an older plugin version, add -Dskiko.renderApi=SOFTWARE to your IDE VM options (Help > Edit Custom VM Options).

See the Troubleshooting guide for more details.

Where do I report bugs or request features?​

Open an issue on GitHub or start a discussion in GitHub Discussions.