Circuit-Sathi

BHAVESH — Sat, 03 Oct 2026 16:41:59 +0000

What I Built
I built Circuit Sathi, an offline, private study companion created for my engineering classmate. During technical exam preparation (like studying MOSFETs, Op-Amps, and small-signal models), revising concepts through active recall is difficult without practice questions.

Existing online quiz tools either require a constant internet connection, distract with notifications, or force students to upload study notes to third-party cloud servers. Circuit Sathi solves this by running completely offline directly in the browser. A student simply pastes their technical lecture notes, selects the number of questions, and instantly receives tailored MCQs with explanations.

Demo
Since Circuit Sathi runs 100% locally on the device to preserve privacy and enable offline study, here is the complete flow:

Initial Interface:

Pasting Technical Notes:

Generated Practice MCQs & Real-time Scoring:

(Optional video demo link if recorded)

Code
The entire project is open-source and structured without dependencies or heavy build frameworks:

{https://github.com/Hitman-886/-Circuit-Sathi-.git}

(Example: [https://github.com/Hitman-886/-Circuit-Sathi-/tree/main]

Quick Start:
Install Ollama.

Download the model:

Bash
ollama pull gemma3:4b
Set the environment variable OLLAMA_ORIGINS to * and restart Ollama.

Open index.html in your browser.

How I Built It
Model & Local Inference: Powered by Google's open-weight Gemma 3 4B model running locally through Ollama via its OpenAI-compatible endpoint (http://localhost:11434/v1).

Frontend Architecture: Built using clean, single-file Vanilla HTML, CSS, and JavaScript. No Node modules, no build steps, and zero external dependencies.
Assistance: Iterated the prompt design and API migration with assistance from Antigravity and Claude.
Why Does Open Innovation Matter?
Open innovation and open-weight models like Gemma 3 enabled key features that closed APIs could not:
True Privacy: Technical notes and study materials remain on the laptop and never hit an external cloud server.
Zero API Tolls: Students can run hundreds of revision quizzes without monthly subscriptions or token billing limits.
Resilience: Operates smoothly during hostel Wi-Fi outages and offline study sessions.
Real Learnings & Technical Trade-offs:
Running Gemma 3 4B locally on consumer hardware (8 GB RAM) gave practical engineering insights:
Inference Speed: Generating 3 MCQs takes around 25–40 seconds on local hardware.

JSON Formatting: While Gemma 3 4B grasps engineering concepts well, compact local models occasionally omit closing brackets when enforcing strict JSON schemas. Handling these edge cases gracefully with client-side retry feedback was a key real-world takeaway.

Prize Categories
Best Use of Gemma (Fully offline, open-weight Gemma 3 4B running on consumer hardware via Ollama).****

DEV Community: BHAVESH

Circuit-Sathi