Advanc

Published using Google Docs
Report abuseLearn more
Coding assignment - Fullstack developer
Updated automatically every 5 minutes
GitHub Issue Analyzer with Local Caching + LLM Processing
Overview
Build a small service with two endpoints that can:

Fetch and locally cache GitHub issues from a repository
Analyze the cached issues using a natural-language prompt and an LLM
You have two days to finish.
You are encouraged to use AI coding tools like Cursor, Claude Code, ChatGPT, etc.

Please note that we expect the coding assignment to be in a working state. In the first round, we'll do a live demo and testing of this coding assignment.

Please share the details here once you have completed the assignment - https://forms.gle/u4QtfALtdSgQhoZc6

What You Need to Build
You will build a server (any language or framework is fine) that provides the following:

1. Endpoint: POST /scan Purpose Fetch all open issues from a given GitHub repository and cache them locally.

Request format
{

"repo": "owner/repository-name"

}

Expected behavior
Fetch all open issues from the GitHub REST API.
Extract and store at minimum:
id
title
body
html_url
created_at
Cache these issues locally using one storage approach of your choice:
Allowed storage options
You must choose one of:

In-memory storage
Fast, simple
Clears when server restarts
JSON file storage
Easy to inspect
Slower for large datasets
SQLite
Most durable
Slightly more setup
Document your choice and reasoning briefly in the README.

Response
Return a summary:

{

"repo": "owner/repository-name",

"issues_fetched": 42,

"cached_successfully": true

}

1. Endpoint: POST /analyze Purpose Take a repo name and a natural-language prompt, retrieve cached issues for that repo, and analyze them using an LLM.

Request format
{

"repo": "owner/repository-name",

"prompt": "Find themes across recent issues and recommend what the maintainers should fix first"

}

Expected behavior
Look up cached issues for the given repo
Combine the prompt + cached issues into an LLM request
Let the LLM generate the analysis (no keyword classification; fully natural-language interpretation)
Return the LLM’s output in a readable response
Requirements
You may use any LLM provider or local model
Keep context size in mind (chunking is acceptable if needed)
Handle edge cases such as:
Repo not yet scanned
No issues cached
LLM errors
Response example
{

"analysis": ""

}

No UI is needed; this is strictly a backend task.

1. Deliverables Please submit the assignment details here - https://forms.gle/u4QtfALtdSgQhoZc6:

A public GitHub repository containing:
Source code
README with:
How to run the server
Why you chose the specific local storage option
A list of prompts you used while building this project, including: (you can add this in README)
Prompts you sent to AI coding tools for help
Prompts used to design the code, fix errors, or generate logic
Prompts used for constructing the final LLM request inside the analyze endpoint (They don’t have to be perfect; this is for understanding your workflow.)