Description
Nexora Context is a tool that can be conected via MCP to LLM's. It gives them context from files, text and URL's the user has uploaded. It saves the user from having to upload the documents into the chat every time.
Our system is a powerful Retrieval-Augmented Generation (RAG) platform designed to handle diverse data sources seamlessly. Users can upload files, plain text, and even URLs. All inputs are processed and stored in three complementary formats:
- Plain text
- Knowledge graph (stores and structures entities, relationships, and properties)
- Vector embeddings
When a query is made, the system performs multi-layered retrieval:
- The query is rephrased twice, enabling three distinct semantic searches in the vector database.
- The knowledge graph is simultaneously scanned for matching entities, their properties, and their relationships.
- Connected entities are re-queried in the vector database to confirm contextual accuracy.
Each retrieval process is isolated to prevent information from one file interfering with another. The collected context is then unified and deduplicated into clean text, which is passed to the user’s LLM. This ensures the model works with high-quality, transparent, and reliable information.
Our system is a powerful Retrieval-Augmented Generation (RAG) platform designed to handle diverse data sources seamlessly. Users can upload files, plain text, and even URLs. All inputs are processed and stored in three complementary formats:
- Plain text
- Knowledge graph (stores and structures entities, relationships, and properties)
- Vector embeddings
When a query is made, the system performs multi-layered retrieval:
- The query is rephrased twice, enabling three distinct semantic searches in the vector database.
- The knowledge graph is simultaneously scanned for matching entities, their properties, and their relationships.
- Connected entities are re-queried in the vector database to confirm contextual accuracy.
Each retrieval process is isolated to prevent information from one file interfering with another. The collected context is then unified and deduplicated into clean text, which is passed to the user’s LLM. This ensures the model works with high-quality, transparent, and reliable information.
Screenshots
