What I Built
I built GhostLog, an AI-powered error log investigator for developers.
Users can paste Python or JavaScript errors and receive an AI-generated investigation containing the error summary, root cause, explanation, step-by-step solution, corrected code example and prevention tips.
Code
GitHub Repository:
https://github.com/karthikeyan24-kk/GhostLog
How I Built It
I used Python, Flask, HTML, CSS, JavaScript, Ollama and Gemma 3 1B.
The Flask backend receives the error logs and sends them to the locally running Gemma model through Ollama's API.
The model analyzes the error and returns a structured explanation.
I also implemented input validation, connection error handling, timeout handling and automated tests.
Why Does Open Innovation Matter?
I chose Gemma 3 1B because it is an open-weight model that can run locally using Ollama.
This means users can analyze their programming errors without sending their logs to an external AI API.
Once the model is downloaded, local inference can work without an internet connection.
Open innovation gives developers the freedom to experiment with models, understand how their AI tools work and build applications without depending on paid cloud AI services.
The Spirit of It
Every developer encounters mysterious errors that seem to appear from nowhere.
GhostLog investigates those hidden problems and helps developers understand the ghosts inside their code.
My goal is to make debugging easier, especially for beginners who are still learning programming.
Prize Categories
Best Use of Gemma
Top comments (1)
Using a small local model for error investigation is a good privacy and cost trade-off. I would evaluate it on noisy real incidents and measure whether its suggested next check reduces time-to-diagnosis, not just whether its summary sounds plausible.