Ask complex biomedical questions in natural language. Get structured, citation-backed answers powered by the IEKB knowledge base and multi-phase AI agent pipeline.
Agent Phases
End-to-end pipeline
Scoring System
Multi-dimensional ranking
Streaming
Real-time token output
How it works
The QA agent connects to IEKB, resolves gene targets, retrieves evidence from multiple tables, and synthesizes a grounded answer.
Multi-phase agent pipeline: analyze, resolve, retrieve, evaluate, rank, synthesize
10-dimension candidate scoring with evidence breakdown
Streaming responses with real-time progress tracking
Capabilities
A purpose-built question-answering system designed for inner ear research — from druggable target discovery to gene-phenotype exploration.
Ask complex biomedical questions in natural language. Compatible with OpenAI, DeepSeek, and any OpenAI-compatible API.
A 10+ phase pipeline: question classification, target resolution, evidence retrieval, batch evaluation, ranking, and synthesis.
Every answer is backed by data from the IEKB database — gene associations, interactions, literature, and more.
Genes scored across 10 dimensions including disease association, mechanism clarity, evidence level, and network centrality.
Optionally include non-probabilistic candidate rankings and deterministic relation support for hypothesis generation beyond known annotations.
Watch the agent think in real time — every phase, every step, every token streamed to your browser via SSE.
Live Demo
Watch the QA agent process a real question — from target resolution through evidence retrieval to the final ranked answer. This is a recorded session replayed in real time.
Get started
Download the complete QA system, unzip, and launch with a single command. The package includes both the Python backend and the Next.js frontend.
unzip iekb-qa.zip && cd iekb-qa
# Linux / macOS
./start.sh
# Windows PowerShell
.\start.ps1Quick steps
iekb-qa.zip./start.sh or .\start.ps1http://localhost:3000