PinkDot AI develops and maintains EscherGraph, an open-source software project. EscherGraph likely provides tools or frameworks for AI-related tasks, though the specific functionality requires further clarification.
Funding
Funding not disclosed
Founders
Product
Problem
Organizations struggle to efficiently extract and analyze key information from unstructured documents, leading to wasted time and missed insights. Existing methods often require manual effort or lack the ability to connect related concepts across multiple documents. This makes it difficult to build a comprehensive understanding of complex topics and answer nuanced questions.
Solution
PinkDot AI offers EscherGraph, an open-source Python package that automatically constructs knowledge graphs from unstructured documents, enabling advanced search and question-answering capabilities. EscherGraph leverages large language models (LLMs) to identify entities, relationships, and key concepts within documents, and then represents this information in a structured graph format. Users can then perform local and global RAG (Retrieval-Augmented Generation) searches to extract relevant information and insights. The platform also provides visualization tools, including interactive plots and dashboards, to explore the knowledge graph and gain a deeper understanding of the underlying data.
Target Audience
EscherGraph is designed for data scientists, researchers, and analysts who need to extract and analyze information from large volumes of unstructured text data.
Features
- Automated knowledge graph construction from documents using LLMs
- Local RAG search for context-aware answers based on graph nodes, edges, and properties
- Global RAG search for high-level insights and conclusions across multiple documents
- Interactive visualization tools, including dashboards and plots, for graph exploration
- Compatibility with OpenAI's GPT4o and text-embedding-3-large models, as well as Jina AI's jina-reranker-v2-base-multilingual
- Flexible API key management via environment variables or keyword arguments
- Support for various file types, including PDFs and text files