product_requirements

🏠 Home

Gemini CLI Bridge - Problem Statement

The Challenge

I recently had PRK surgery and need to reduce eye strain while coding. I'm using Google's Gemini CLI as an agentic code mentor, but its terminal interface creates too much context switching and strain. I want to work entirely within VSCode—writing prompts in one file, getting responses in another—both located inside my chosen repository, all in a comfortable, readable format.

What I Want

A background service that watches my prompt.md file (in my repository) and automatically feeds it to Gemini CLI, then saves the response in TUTORIALS/output.md (in the same repository), where I can read it comfortably.

My Ideal Workflow

  1. Open prompt.md (inside my repo) in VSCode and type my question
  2. Save the file (Ctrl+S)
  3. Bridge detects the save and sends content to Gemini CLI running in background
  4. Gemini processes and responds
  5. Response gets written to TUTORIALS/output.md (also inside my repo) in pure Markdown
  6. I manually open the output file in VSCode preview or Chrome to read it
  7. Repeat for next prompt

Key Requirements

Must Have:

Must Not Have:

Safety:

Current Problem

The v0.0.4 I have now:

See actual codebase

Success Looks Like

I can work for hours just switching between two VSCode tabs—prompt.md for writing prompts, TUTORIALS/output.md for reading responses. No terminal windows, no external browsers (unless I choose), no auto-updating content. Just clean, readable Markdown responses that help me understand my codebase better while being gentle on my recovering eyes.

The bridge runs silently in the background, I forget it exists, and it just works every time I save a prompt.