This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
What I Built
Who I Built This For: My friend, who is currently enrolled in a fast-paced coding bootcamp.
The Problem & Solution: When bootcamp students get stuck, they often paste errors into ChatGPT and get the final code—bypassing the learning process completely. To help my friend actually learn, I built a Socratic "Rubber Duck" tutor using Backboard's Stateful Vector Memory and Semantic RAG. Instead of just giving the answer, it acts as a patient tutor, retrieving relevant coding concepts and asking guiding questions while tracking their progress across sessions.
Demo
When my friend asks for help with a math error in their code, the agent uses a live Backboard RAG citation to guide them rather than writing the fix:
🧠 Backboard RAG: Concept: Mean Squared Error (MSE)... (Score: 0.53)
"Let's look at how you are calculating squared_errors. What happens mathematically when you subtract targets ** 2 from the differences, rather than squaring the differences themselves?"
Code
aditya-prog-bit
/
rubber-duck-companion
Socratic Coding & Debugging Companion built with Backboard R-CLI and Vector Memory RAG for Hacktoberfest 2026
🦆 The "Rubber Duck" Coding Companion
Built with Backboard R-CLI & Socratic Agent Harness for Hacktoberfest 2026
"Ship something that solves a real problem for a friend or someone you love."
🎯 Overview
The Rubber Duck Coding Companion is an interactive, Socratic AI debugging companion built to help beginners and friends learn how to code and debug on their own—without generic AI immediately spoiling the answer.
Generic AI models usually rewrite the whole function the moment you paste an error. While that gets past the compiler, it deprives learners of the essential "aha!" moment.
By leveraging Backboard's Vector Memory & RAG Infrastructure along with its custom agent harness, we built a stateful debugging companion that:
- Requires NO paid LLM chat tokens: Runs entirely on Backboard's free vector memory tier.
- Stores persistent student memories: Logs student learning milestones, mistakes, and eureka moments across sessions.
- Retrieves pedagogical…
How I Built It
Socratic Sub-Agent: I configured .backboard/agents/rubber_duck.md with strict rules: NEVER write the final code. Ask guiding questions one at a time.
Vector RAG Knowledge Base: I loaded Backboard's vector database with debugging concepts. Backboard's semantic search retrieves the right hints without using paid LLM chat tokens.
Persistent Student Memory: When my friend solves a bug, Backboard automatically saves a persistent cloud memory, remembering their achievements for future sessions.
Memory & RAG Studio UI: I built a local dashboard at 127.0.0.1:8000 where my friend can inspect their live vector memories and test semantic searches.
Why Does Open Innovation Matter?
Closed chatbots force you into their default "eager-to-solve" behavior. Open innovation frameworks like Backboard allowed me to strip that out and inject a strict pedagogical framework tailored to how a student actually needs to learn. Furthermore, utilizing accessible vector memory locally in their terminal ensures my friend's bootcamp assignments remain completely private.
Prize Categories
- Best Use of Backboard
Top comments (0)