Echo: Local AI Empathy Assistant

Inspiration

In today's remote-first world, text communication is a minefield. Without tone of voice or body language, a quick message can easily be misinterpreted as rude or aggressive. We've all sent emails we regretted or struggled to phrase difficult feedback. We wanted to build a tool that acts as a "second brain" for emotional intelligence one that helps us connect better, without sacrificing privacy by sending personal drafts to the cloud.

What it does

Echo is a "Vibe Check" for your messages. You paste a rough or emotional draft (e.g., an angry email to a coworker), and Echo uses a local Large Language Model (LLM) to analyze the sentiment. It then rewrites the message to be professional and empathetic while preserving the original intent. Users can toggle between different tones Calm, Assertive, or Empathetic and get an explanation of why the AI's version works better.

Our Team

Kamil Piejko - Frontend Designer Mateusz Grabowski - Prompt Engineering and structure

Challenges we ran into

The biggest challenge was the Browser2Localhost connection. Modern browsers block requests from a web page to local servers due to CORS (Cross-Origin Resource Sharing) policies. We had to configure the Ollama server specifically to accept requests from our Vite application (OLLAMA_ORIGINS="*") and implement robust error handling in React to manage the connection state with the local AI model.

What we learned

We learned a lot about Prompt Engineering for JSON output making sure the model follows strict formatting rules even when the user input is chaotic. We also gained deep experience in bridging the gap between modern web frameworks and local system level AI tools.

Built With

• mistral`
• ollama`
• radix
• react`
• tailwind-css`
• typescript`
• vite`