> For the complete documentation index, see [llms.txt](https://docs.biobox.io/guide/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.biobox.io/guide/how-to/ask-multi-modal-questions/ask-questions-with-natural-language-graphrag.md). # Ask questions with Natural Language (GraphRAG) ## What is GraphRAG? **GraphRAG (Graph-Retrieval Augmented Generation)** is a system that lets you ask natural-language questions about your data — and get accurate answers directly from your custom knowledge graph. Instead of manually writing query language or browsing complex graph structures, GraphRAG uses an AI agent to: 1. **Understand your question** — It interprets what you’re asking in plain English (e.g., *“Which genes are upregulated in ALS?”*). 2. **Translate it into a graph query** — The AI automatically generates the correct Cypher query based on BioBox’s custom ontology and data model. 3. **Retrieve and summarize results** — The query runs against your knowledge graph, and GraphRAG summarizes the findings into clear, readable text alongside the supporting evidence. This approach combines the **reasoning ability of large language models** with the **accuracy and structure of a knowledge graph**, allowing scientists to explore complex biological relationships without needing to know query syntax. ## Asking Questions with Natural Language Ensure that your knowledge engine is toggled to "Natural Language". Type your question into the search bar and hit enter. {% embed url="" %} When you submit a question, the **question** and the **agent’s interpretation** are pinned to the top of the page for easy reference. The agent then begins transforming your natural-language question into a **graph query**, which you can observe in real time under the **Prompts** tab. Once the answer has been generated, the interface automatically switches to the **Results** tab. The results section is composed of several specialized agents, each providing a different perspective on the retrieved data (e.g., Biological Concept Agent, Graph Paths Agent, Text-Based Agent). For example, the question **“Which genes are upregulated in ALS?”** is converted into the following graph query: {% code overflow="wrap" %} ``` (d:Disease {displayName = "familial amyotrophic lateral sclerosis"})<- studies condition<- (d2:DiffExpDataset)-> upregulates-> (g:Gene) ``` {% endcode %} #### Results Summary Agent The agent provides a text-based summary of all of the data returned in the search. Select "View More" next to each agent summary to view a longer, comprehensive summary. {% embed url="" %} #### Text-Based Agent The **Text-Based Agent** retrieves relevant **electronic laboratory notebooks (ELNs)** that contain information related to your question. If ELNs are integrated into your knowledge graph, this agent will automatically display all notebooks containing supporting data. A **summary** of the findings appears above the list of ELNs, and selecting a specific notebook opens a **side panel** where you can view its detailed content. {% embed url="" %} #### Biological Concept Agent The **Biological Concept Agent** retrieves all biological concepts and their related data objects that help answer your question. The concepts and data returned are specific to your **custom knowledge graph**. For example, the question **“What genes are upregulated in ALS?”** is automatically transformed into a graph query involving the biological concepts **Genes**, **Disease**, and **Differential Expression Datasets**. You can explore the results by selecting specific concepts from the **left-hand panel**. The corresponding data objects for each concept will appear in the **table view** on the right. {% embed url="" %} #### Graph Paths Agent The **Graph Paths Agent** retrieves the **individual graph paths** that connect concepts and provide evidence for your question. This view offers additional flexibility by allowing you to **customize which data properties are displayed** and **how they are ordered**. Use **Edit columns** to choose which data properties to show, and **Reorder columns** to drag and arrange them in your preferred order. For instance, in the question **“Which genes are upregulated in ALS?”**, a graph path may include the specific dataset, gene symbol, log₂ fold change, and adjusted p-value associated with each result. {% embed url="" %}