I built a runbook generator with FastAPI and Groq — here's how it works

Every time there's an incident or a deployment, I need a runbook. The structure is always the same: Overview → Prerequisites → Procedure → Verification → Troubleshooting → Rollback.

I got tired of filling in the same skeleton over and over, so I built Auto Runbook Generator.

What it does

Paste a Kubernetes YAML, a docker-compose file, or just describe your system in plain English. Pick a runbook type (deployment, incident response, rollback, maintenance, onboarding). Get a complete Markdown runbook back in seconds.

Live demo: https://auto-runbook-gen-v0-1-0.onrender.com

How I built it

Backend: FastAPI + AsyncGroq

The core is a single /api/generate endpoint. It takes the user's input and runbook type, builds a prompt, and calls Groq's API with llama-3.3-70b.

response = await get_groq().chat.completions.create(

    model="llama-3.3-70b-versatile",

    messages=[

        {"role": "system", "content": SYSTEM_PROMPT},

        {"role": "user", "content": user_prompt},

    ],

    max_tokens=2048,

    temperature=0.3,

)

I use AsyncGroq (not Groq) because FastAPI is async — calling a sync client inside an async endpoint blocks the event loop.

Rate limiting: slowapi

Free tier gets 10 generations/day per IP. Pro API key bypasses this.

@app.post("/api/generate")

@limiter.limit("10/day")

async def generate(request: Request, ...):

The system prompt matters a lot

The prompt enforces a consistent structure with 6 sections every time:

# [Service Name] Runbook

## Overview

## Prerequisites

## Procedure

## Verification

## Troubleshooting

## Rollback

Low temperature (0.3) keeps outputs consistent and factual rather than creative.

Frontend: zero dependencies

Single HTML file, vanilla JS. fetch() to call the API, navigator.clipboard to copy the result. No React, no build step.

Deploy: Docker on Render

FROM python:3.12-slim

WORKDIR /app

COPY requirements.txt .

RUN pip install --no-cache-dir -r requirements.txt

COPY . .

CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000"]

Results

The output quality is good enough to use directly for most common cases. Kubernetes and Docker scenarios work especially well since the LLM has strong training data for those.

Source

GitHub: https://github.com/Jedsadakorn-Suma/auto-runbook-gen

Self-hostable (MIT). Needs a free Groq API key from console.groq.com.