DEV Community

Thabasvini
Thabasvini

Posted on

Bug Whisperer – AI-Powered Debugging with Redis

#redischallenge #devchallenge #database #ai

Redis AI Challenge: Real-Time AI Innovators

This is a submission for the Redis AI Challenge: Real-Time AI Innovators.

What I Built

Bug Whisperer is an AI-powered debugging assistant that listens for error logs in real time, classifies their severity, explains the cause, and suggests fixes.
It stores embeddings of each bug so it can instantly recall and provide suggestions for similar issues in the future.

Key highlights:

  1. Real-time bug ingestion via Redis Streams

  2. AI-powered suggestions using Google FLAN-T5

  3. Semantic similarity search via Redis Vector Store + Sentence Transformers

  4. Template fallback for common errors

  5. Streamlit dashboard to view recent bugs & chat with the assistant

Demo

GitHub Repo: https://github.com/Thabasvini/Bug-Whisperer-AI-Powered-Debugging-with-Redis-Vector-Search

dashboard1

dashboard2

dashboard3

How it works in action:

Producer/Simulated App pushes an error into Redis Stream bug_logs

Consumer reads new error → classifies severity → checks for similar bugs in Redis → generates suggestion

Dashboard displays recent errors + lets users ask about new bugs directly

How I Used Redis 8

I leveraged multiple Redis 8 AI-focused features:

  • Redis Streams: Acts as the real-time log ingestion layer (bug_logs).

  • Redis Hashes & Lists: Store structured bug reports (bug:{id}) and maintain an ordered list of recent bugs (bug_index).

  • Redis Vector Store: Store semantic embeddings of bug messages from all-MiniLM-L6-v2.

  • Enables semantic recall for similar bugs with high cosine similarity.

  • Semantic Caching: When a bug is seen again, the system instantly retrieves a stored suggestion instead of calling the model again, reducing cost & latency.

Solo project by @aibythabasvini

Top comments (0)