Skip to content

JetBio

Is a reimagined IDE for computational biology that treats biological correctness as a first-class debugging target.

Code Video Cerebral Valley

Core trick

Stress-tests bioinformatics pipelines with synthetic biological edge cases before execution, then traces bad results back to root-cause steps.

Tags

sciencedev tools

Description

Is a reimagined IDE for computational biology that treats biological correctness as a first-class debugging target. Utilizes ConTree's pipeline branching, it replaces the broken silence of bioinformatics failures — where pipelines complete successfully but produce wrong science — with three AI agents: BioTrace (traces a bad result back to its root cause), BioFuzz (stress-tests pipelines with synthetic biological edge cases before they run), and BioReview (reviews analysis code the way a peer reviewer would). The IDE surface shows a live pipeline branch tree, a streaming debug console, and an auto-generated methods section — all in one canvas.
AUW160150/JetBrain★ 00 forksTypeScriptCreated Apr 19, 2026 · last push Apr 19, 2026

README excerpt

# JetBio — BioTrace Debug Agent

> AI-powered bioinformatics debugging. Paste your broken DESeq2 / RNA-seq / scRNA-seq analysis, get structured root-cause reasoning and a concrete, runnable fix.

![JetBrains-style dark UI](https://img.shields.io/badge/UI-JetBrains%20style-4D9FE0?style=flat-square)
![TypeScript](https://img.shields.io/badge/TypeScript-5.5-3178C6?style=flat-square&logo=typescript)
![React](https://img.shields.io/badge/React-18-61DAFB?style=flat-square&logo=react)
![OpenRouter](https://img.shields.io/badge/LLM-OpenRouter-FF6B35?style=flat-square)

---

## Demo

Paste this into the Debug Console:

```
My DESeq2 analysis returned 0 significant genes. Volcano plot is flat.
I ran: DESeq2 with design = ~ condition, comparing treated vs control, n=6 samples per group.
```

**What happens:**

1. **BioTraceAgent** runs 5 systematic reasoning steps (symptoms → dispersion → design matrix → metadata → batch effects)
2. **DiagnosticAgent** distills those steps into a single root cause with confidence score + evidence
3. **FixAgent** generates a concrete, runnable R fix

**Expected output:**

| Panel | Content |
|---|---|
| Reasoning Chain | 5 gutter-annotated steps with observations + inferences |
| Root Cause (red) | "Design formula missing batch covariate" — 92% confidence |
| Fix (green) | `design = ~ batch + condition` with full R code + ordered steps |

---

## Architecture

```
┌──────────────────────────────────────────────────────────────┐
│                   Browse

Team