🚀 ContextStream – Eliminate Context Debt

Extract Decisions • Rationale • Next Steps from Any Conversation

Live App: https://context-stream-lilac.vercel.app

Backend API: https://contextstream.onrender.com/api/extract

📌 Overview

Teams today lose hours every week rewriting and repeating information buried in Slack threads, meeting notes, and emails. This leads to:

Missed decisions

Lost rationale

Slow onboarding

Repeated explanations

Decision reversals

This invisible productivity drain is Context Debt.

ContextStream eliminates context debt by automatically extracting:

🟦 Decisions

🟩 Rationale

🟨 Next Steps

from any conversation pasted into the interface.

This prototype is built for Startup School: Prompt to Prototype.

🎯 Key Features

✔ Mock AI Extraction Engine

Extracts structure instantly using heuristic rules (no API keys needed).

✔ Live Deployed App

Frontend on Vercel

Backend on Render

Always-available public URLs for judges

✔ Lightweight Frontend

HTML + CSS + Vanilla JS for speed and simplicity.

✔ Modular Backend

Node.js + Express API ready to be swapped with Gemini/OpenAI in future versions.

✔ Zero Setup

Runs instantly in the browser. Perfect for demos, evaluations, and rapid iteration.

🧪 Try It Yourself

Open the live app

Paste any conversation like:

We will finalize the landing page by Friday.
I will send 3 design drafts because marketing asked for options.
Next: push wireframe today.
Action: run A/B test on hero headline.

Click Generate Context

See structured insights extracted instantly.

🛠 Tech Stack

Frontend

HTML

CSS

JavaScript

Vercel deployment

Backend

Node.js

Express

CORS

dotenv

Render deployment

Architecture Diagram

Frontend (Vercel)
     |
     | POST /api/extract
     v
Backend (Render: Node + Express)
     |
     v
Mock Extraction Engine → Structured Output (JSON)

📂 Project Structure

ContextStream/
│
├── server/
│   └── index.js               # Mock extraction backend (Express)
│
├── frontend/
│   ├── index.html             # UI
│   ├── styles.css             # Styling
│   └── app.js                 # API calls + rendering
│
├── docs/                      # Submission documentation (optional)
│
├── package.json
└── README.md

🚀 How to Run Locally

1. Clone repo

git clone <repo-url>
cd ContextStream

3. Install backend dependencies

npm install

4. Run backend

npm start

Backend will run on:

http://localhost:3000

4. Open frontend manually Just open:

frontend/index.html

in any browser.

🟨 Golden Prompt (Used to Start the Project)

You are an AI system that extracts “Context Insights” from team conversations.

Given any conversation (Slack thread, meeting notes, or emails), identify and separate:

1. "decisions" — what was agreed or committed
2. "rationale" — reasons behind the decisions
3. "next_steps" — action items or tasks assigned

Return output ONLY in the following JSON format:

{
  "decisions": [],
  "rationale": [],
  "next_steps": []
}

The purpose is to eliminate “context debt” by making key information instantly accessible.

🚧 Future Roadmap

Replace mock engine with Gemini 1.5 Flash / Pro

Slack integration

Gmail thread extraction

Timeline of decisions with versioning

Team workspace + authentication

Analytics dashboard for “context health”

📜 License

MIT License.

🙌 Acknowledgements

Built as part of Google’s Startup School – Prompt to Prototype program.