• 

  Data Ingestion Button

• 

  Data Ingestion Workflow in Action

• 

  JSON Representation of Recorded Actions

• 

  Chat Output

• 

  Chat Output with Option to view relevant Screenshots

• 

  Screenshot Viewer

Caveat Bot: Automate Your Knowledge Base Creation

Have you spent hours trying to set up configurations for a new project? Or maybe it's a fresh start on a new computer for a new job? These setups can take forever, and documenting what you've tried, what failed, and what finally worked is a nightmare to do efficiently.

Enter Caveat Bot.

Caveat Bot is a game-changing VS Code extension that automatically tracks your every move: commands and their outputs, code changes, screenshots, and even your notes. It leverages Qroq's high-speed AI to build a multimodal knowledge base, giving you fast, intelligent access to your past work without wasting time documenting manually. With Caveat Bot, you can avoid repetitive mistakes, streamline your workflow, and stay organized like never before. 🧠💻

Project Description

Inspiration

I’ve become obsessed with building a personal knowledge base for general things. I’m a big fan of Notion and have invested a lot of time creating and organizing my data there. However, I realized I didn’t need to spend all that time building it manually. Caveat Bot was born out of the idea of automating this process to save time and effort.

What It Does

Caveat Bot captures your commands, code changes, notes, and screenshots, then uses advanced AI to transform them into a searchable, organized knowledge base.

How It Was Built

I captured all actions using VS Code's Extension API and tooling functionality and stored them as JSON. These actions were ingested with Groq using LangChain, orchestrating a whole variety of prompts. Then, I used OpenAI embeddings and Pinecone to store embeddings of different modalities along with metadata for later use. I also set up a chat interface, which executed an agentic workflow that retrieved data relevant to the task and presented the results, along with images that could be useful.

Challenges I Ran Into

Creating a VS Code extension for the first time—especially solo—was a steep learning curve. Extensions are notoriously hard to develop, but persistence paid off.

Accomplishments That I’m Proud Of

I’m proud to have a working product that integrates seamlessly into VS Code and is already somewhat user-friendly, offering real value to developers.

What I Learned

I learned not to underestimate the effort required to build a clean, functional UI. Additionally, I gained valuable insights into designing efficient query mechanisms and orchestrating prompt chains using Qroq.

What’s Next for Caveat Bot

The next step is to productionize the extension and release it to the public. I aim to enable users to add their own Qroq, Pinecone, and OpenAI API keys for embeddings, making the tool highly customizable and accessible.

Built With

• groq
• javascript
• langchain
• openai
• pinecone
• typescript