Facilitating the Spread of Knowledge and Innovation in Professional Software Development

Write for InfoQ


Choose your language

InfoQ Homepage News The Devoxx Genie IntelliJ Plugin Provides Access to Local or Cloud Based LLM Models

The Devoxx Genie IntelliJ Plugin Provides Access to Local or Cloud Based LLM Models

Devoxx Genie, a 100% Java-based JetBrains IntelliJ IDEA Plugin, uses local- or cloud-based Large Language Models (LLMs) for generating unit tests and explaining, reviewing and improving source code.

The plugin supports local LLM providers Ollama, LM Studio and GPT4All. Local LLMs provide a good solution for companies who are hesitant to send source code to the cloud. Alternatively, the following cloud-based LLM providers may be used: OpenAI, Mistral, Anthropic, Groq and DeepInfra.

Stephan Janssen, founder of Devoxx(4kids), introduced Devoxx Genie in April 2024.The plugin requires a minimum of JDK 17 and IntelliJ IDEA 2023.3.4.

The plugin may be installed via the JetBrains Marketplace website or inside IntelliJ via Settings | Plugins and then selecting DevoxxGenie:

The plugin requires at least one URL for a local LLM or an API key for a cloud-based LLM. However, it's also possible to configure multiple models at the same time. The configuration is accessible via Settings | Devoxx Genie Settings:

Finally, after selecting some source code, the plugin's panel can be used with one of the /help, /test, /review, /explain and /custom prompts:

Unfortunately, using an LLM might result in LLM hallucinations when answers are coherent and grammatically correct, but factually incorrect or nonsensical. Retrieval Augmented Generation (RAG) might be used to improve the results of traditional LLMs by using a vector database or a feature store to augment traditional LLMs by providing the prompt with context.

Janssen also introduced RAG Genie in April 2024, an LLM RAG prototype to evaluate embeddings and splitting strategies. The tool provides detailed information in a graphical user interface, for example on where sentences are split. Janssen has future plans to include RAG within the Devoxx Genie plugin for context and questions. RAG also allows the orchestration of models based on the price and quality of the answers. This makes it possible to use a cheaper model for less critical answers and a pricier model for the more critical answers.

Janssen explained RAG Genie and Devoxx Genie in his session at Devoxx France.

About the Author

Rate this Article