Skip to content

KataLab

KataLab is a human-in-the-loop GenAI Agent system built with Google AI tools to solve costly and slow company onboarding challenges.

Code Video Cerebral Valley

Core trick

An agent reads company documents and profiles, then generates a custom kata repository for the new hire with human review.

Tags

educationagents

Description

KataLab is a human-in-the-loop GenAI Agent system built with Google AI tools to solve costly and slow company onboarding challenges. It receives company documents and employee profiles to generate custom, production-grade coding challenges (Katas) that serve as tailored learning tools for the new employee. This solution accelerates new engineers' time-to-first-contribution, ensuring they are productive and integrated faster than traditional training allows.
perelloliver/katalab★ 00 forksPythonCreated Dec 13, 2025 · last push Dec 14, 2025

README excerpt

# Katalab


<img width="1379" height="786" alt="Screenshot 2025-12-13 at 17 47 10" src="https://github.com/user-attachments/assets/b3732c09-dbaa-4b58-adfa-3dc930bab3a4" />

Turn company documents into coding katas for custom employee team onboarding.

Built entirely with Antigravity and Google AI Studio for the DeepMind x Cerebral Valley LDN hack.

**We had fun building and liked the idea, so this repo is currently being taken from hack to MVP.**

## Overview

Katalab parses company context (CVs, engineering blogs, job specs) and uses an Agent to design and generate a custom "Coding Kata" repository tailored to that company's stack and values.

## Structure

- **Backend**: FastAPI application (`src/backend`)
- **Frontend**: Vanilla JS/CSS web interface (`src/frontend`)
- **Agent**: Gemini powered planning and generation.

## How to Run

The frontend is served statically by the backend, so you only need to run the Python application.

1.  **Install Dependencies**:
    ```bash
    uv sync
    ```

2.  **Set Environment Variables**:
    Ensure you have a `.env` file with `GOOGLE_API_KEY`.

3.  **Run the App**:
    ```bash
    uv run uvicorn src.backend.main:app --reload
    ```

4.  **Access the UI**:
    Open [http://localhost:8000](http://localhost:8000).

## Usage

1.  Upload documents.
2.  Review the generated plan.
3.  Provide feedback to refine the plan.
4.  Click "Approve & Build" to generate and download the Kata repository.

Team