02 / Support systems
A knowledge-based AI assistant
Qragy brings knowledge retrieval, conversation and human handoff into a self-hosted application.
Open source · Updated · September 2026
What I worked on
Product workflow, retrieval architecture and implementation with AI development tools.
The starting point
A support assistant needs more than an answer generator: relevant knowledge, a clear conversation flow and a route to a human when the answer is insufficient.
How I built it
Combine full-text and vector retrieval to find relevant knowledge.
Include an admin workspace, conversation state and human handoff in the same application.
Use embedded storage and support multiple model providers, including local models through Ollama.
The result
A public, self-hostable application that brings retrieval, support workflows and evaluation tools together. The repository contains installation instructions and architecture documentation.
Answer quality depends on the source documents, retrieval and chosen model. Hosting and model-provider costs vary with the setup.
Code & resources
Node.js / JavaScript / LanceDB / SQLite / RAG