We built a simple retriever that uses an LLM to generate multiple relevant search queries, executes a search for each query, chooses top links per query, loads information from chosen links into the context window of an LLM for synthesis. The retriever is designed to be easily configurable and can be run in private mode using various tools such as LlamaV2 and GPT4all embeddings. It has the potential to benefit from adding agentic properties, allowing it to ask for more information or construct final answers with multiple "write" agents.