iDC-NEU/YiGraph

What it solves

YiGraph addresses the difficulty of extracting insights from complex relational data for non-technical users. It eliminates the need for users to manually write complex graph queries or understand specific graph algorithms, allowing them to transform natural language business questions into executable and traceable graph analysis reports.

How it works

The system uses the Analytics-Augmented Generation (AAG) framework, which combines Large Language Models (LLMs) with professional graph computing engines. Instead of relying on the LLM to reason through data via text, YiGraph uses the LLM to plan the analysis process and then invokes verifiable, reproducible graph algorithm modules (via NetworkX and Neo4j) to perform the actual calculations. It can automatically extract entities and relationships from raw text, logs, or tables to build a task-specific graph before executing the planned analysis.

Who it’s for

• Business Analysts: Non-technical users who need to identify patterns like money laundering loops or fraud networks without writing code.
• Risk Investigators: Professionals in finance, e-commerce, or supply chain management looking for hidden associations and operational risks.
• Advanced Users/Experts: Those who want to collaborate with the AI to refine analysis workflows or specify exact algorithms for complex investigations.

Highlights

• Extensive Algorithm Library: Includes over 200 built-in graph algorithms across 21 categories (e.g., PageRank, Louvain, Dijkstra).
• AAG Framework: Ensures reliability by using LLMs for planning and professional modules for execution, making results traceable and reproducible.
• Multi-Source Data Support: Automatically constructs graphs from unstructured text, logs, and structured tables.
• Flexible Operating Modes: Offers Normal (fully automated), Interactive (collaborative), and Expert (user-defined) modes.
• End-to-End Pipeline: Covers everything from raw data ingestion and graph construction to final report generation.