https://github.com/andrewdarnall/leetllama
LeetLLama is an AI-powered coding mentor that helps you learn algorithms and programming through interactive, personalized problem-solving—like pair-programming with a senior dev.
https://github.com/andrewdarnall/leetllama
ai developertools edtech leetcode llama3 llm milvus ollama selflearning
Last synced: about 2 months ago
JSON representation
LeetLLama is an AI-powered coding mentor that helps you learn algorithms and programming through interactive, personalized problem-solving—like pair-programming with a senior dev.
- Host: GitHub
- URL: https://github.com/andrewdarnall/leetllama
- Owner: AndrewDarnall
- License: gpl-3.0
- Created: 2024-11-13T10:21:24.000Z (11 months ago)
- Default Branch: main
- Last Pushed: 2025-07-28T00:36:48.000Z (2 months ago)
- Last Synced: 2025-07-28T01:05:08.408Z (2 months ago)
- Topics: ai, developertools, edtech, leetcode, llama3, llm, milvus, ollama, selflearning
- Language: Python
- Homepage:
- Size: 6.35 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# 🦙 LeetLLama
[](https://aws.amazon.com/)
[](https://www.docker.com/)
[](https://kubernetes.io/)
[](https://www.python.org/)
[](https://ollama.com/)
[](https://www.gnu.org/licenses/gpl-3.0.html)
# Welcome to **LeetLLama**
**LeetLLama** is a lightweight yet forward-thinking **proof of concept** (PoC) for an interactive, AI-augmented platform designed to empower aspiring software engineers through self-directed learning.
---
## 🎯 Purpose & Vision
The primary goal of this project—beyond _“reinventing the wheel for academic purposes”_—is to demonstrate a core philosophy in modern tech:
> **Artificial Intelligence should augment, not replace, human developers.**
In this PoC, we explore how **AI-driven tools** can enhance the way programmers learn complex topics like algorithms and data structures, providing a more personalized, adaptive, and contextual learning experience.
---
## 💡 Concept Overview
At its core, LeetLLama enables learners to:
- Interact with a **fine-tuned language model** trained on solved LeetCode-style problems.
- Receive guidance as if from a more experienced developer or mentor.
- Customize their learning journey based on:
- ✏️ **Coding style**
- 🧠 **Skill level**
- 🎯 **Problem preferences** (difficulty, topic, and more)
The goal is to simulate the benefits of a **pair-programming** session with a senior engineer—one who not only knows the answers but can explain _why_ they work, and tailor their advice to your unique thinking process.
---
## 🧱 Tech Stack & Architecture
- 🐍 Python
- 📦 Dockerized microservice architecture
- 🧵 Message queue for asynchronous communication
- 🤖 LLM backend
- 💾 Vector store
- ☁️ Cloud-deployable with Kubernetes
# Software Requirements
| `Software` | `Version` |
|-------------------|-----------|
| `Python` | `3.11` |
| `pip` | `25.1` |
| `Docker` | `28.1.1` |
| `Docker-Compose` | `2.36.0` |
# Hardware Requirements
> These are *minimun* requirements
|`Component` | `Vendor` | `Model` |
|---------------------|-----------|--------------------|
| `CPU` | `Intel` | `8th Gen` |
| `GPU` | `NVIDIA` | `GeForce 4070 RTX` |
| `RAM` | `Any` | `DDR4` |
| `Secondary Memory` | `SSD` | `100 GB` |
# Setup & Configuration
1) Start by *building* the entire project
```bash
docker compose build
```
2) Then *up* the project with
```bash
docker compose up -d
```
3) Connect to `ollama` to *download* the required models:
```bash
llama3.2:3b
qwen3:4b
phi3:3.8b
gemma3:4b
```
Using the commands:
(Connect to the ollama container)
```bash
docker exec -it leetllama_ollama /bin/bash
```
(Pull the models)
```bash
ollama pull llama3.2:3b && ollama pull qwen3:4b && ollama pull phi3:3.8b && ollama pull gemma3:4b
```
Then `exit`
4) Then, download all the `Python` dependencies required to populate the *Milvus* Knowledge Base with the ***Leetcode Dataset***, running this script:
(Download the requirements)
```bash
python -m pip install -r requirements.txt
```
(Populate the Milvus Vector Store)
```bash
cd ~/data
python load_milvus_kb.py
cd ../
```
5) Now you may connect to the *Frontend UI* at the following URL:
```bash
http://localhost:1291/
```
# Some Samples
This is the frontend UI you will see upon ***upping*** the `docker-compose` file.
The gear button aids with the selection of the ***models*** to be used for the ***inference***:
Then once you have chosen the model you seek, ***press the confirm button***
---
1) In this simple example, I asked how I could *invert a linked list*
2) In this more articulated example, I gave the model the entire question for the *infamous* `two sum` probelm
# Disclaimer
This repo is still an early ***PoC*** and not ready for being *forked*, once the `Cloud Setup` and `Better Architecture` will be implemented, then
you may proceed with ***forking*** the repo, and to know for sure, this message will **no longer appear in the README**
---