revela - Data Copilot for the Web

AI-powered data analysis for the web. Analyze tables and charts on any webpage with instant insights and interactive exploration.

Overview

Revela is a Chrome extension with a cloud-native backend that provides ephemeral, privacy-first AI analysis of data visualizations using Google's Gemma 3 models. Simply hover over any table or chart while browsing to get instant insights or start a conversational analysis.

How Revela Helps

• Instant Analysis: Hover over any table or chart on a webpage to see AI-generated insights
• Deep Exploration: Ask questions and perform complex analytics through an interactive chat interface
• Chart Generation: Automatically generate visualizations from data using matplotlib
• Privacy-First: Ephemeral sessions with no persistent storage of user data
• Zero Setup: Works directly in your browser with any public webpage

Technologies Used

• Google Cloud Run: Serverless deployment with auto-scaling (CPU and GPU instances)
• Gemma 3 12B: Google's 12B parameter language model for data analysis
• Ollama: Local LLM inference server
• Flask: Python web framework for API layer
• Polars: High-performance DataFrame library for data processing
• Python: Backend service implementation
• JavaScript: Chrome extension and frontend
• Chrome Extension API: Browser integration

System Components

1. Chrome Extension (Frontend)

The browser-side component that:

• Detects analyzable tables and images on web pages
• Displays hover icons beside data elements
• Provides two interaction modes:
  • Quick Insights: Instant AI summary
  • Deep Analyze: Interactive chat sidebar
• Extracts and sends data to the backend API
• Manages user sessions

2. Service Layer (Backend API)

Flask-based API service running on Google Cloud Run (CPU):

• Manages ephemeral analysis sessions (30-minute timeout)
• Processes HTML tables and images
• Orchestrates LLM queries to Ollama service
• Executes Polars-based data queries
• Generates matplotlib charts
• Auto-cleanup of inactive sessions
• Resources: 2 vCPU, 2 GiB RAM, scales 0-10 instances

3. Ollama LLM Container

Inference service running on Google Cloud Run (GPU):

• Hosts Gemma 3 (12B parameter) model
• Handles vision analysis for charts and images
• Processes natural language queries
• GPU-accelerated inference with NVIDIA L4
• Resources: 4 vCPU, 8 GiB RAM, 1x NVIDIA L4 GPU, scales 0-5 instances

Architecture

The architecture consists of three main components:

1. Chrome Extension - Browser-side detection and UI for seamless integration with any webpage
2. Backend API (Cloud Run - CPU) - Flask service handling session management, data processing with Polars, and code execution
3. Ollama LLM (Cloud Run - GPU) - GPU-accelerated inference with Gemma 3 for vision and language analysis

Cloud Run Advantages

• Auto-scaling: Scales to zero when idle, reducing costs
• Serverless: No infrastructure management required
• GPU Support: Native support for GPU-accelerated inference
• Source Deployment: Deploy directly from source code
• Service Authentication: Built-in IAM for secure service-to-service communication

Use Cases

1. Quick Data Exploration

Hover over a table on Wikipedia, financial sites, or research papers to instantly understand trends, outliers, and key statistics without leaving the page.

2. Comparative Analysis

Ask questions like "Which country has the highest GDP?" or "Compare the top 5 values" to get immediate answers with supporting data.

3. Chart Generation

Request visualizations directly from table data: "Show me a bar chart of GDP by country" to generate matplotlib charts on demand.

4. Image Analysis

Analyze existing charts and visualizations: "What trends does this chart show?" to extract insights from images.

5. Complex Analytics

Perform Polars-based queries on web data: "What's the average GDP of Asian countries?" with automatic data filtering and aggregation.

6. Educational Research

Students and researchers can quickly analyze data tables in academic papers, online courses, or reference materials.

7. Financial Analysis

Analyze stock tables, economic indicators, and financial reports directly from news sites or investor portals.

8. Competitive Intelligence

Extract insights from competitor data tables, market research reports, and industry statistics found online.

Project Setup

Local Development

Prerequisites

• Python 3.11+
• Node.js 16+
• npm
• Chrome browser
• Ollama (for local LLM inference)

1. Backend Setup

cd revela-app

# Create virtual environment
uv venv .venv

# Activate virtual environment
source .venv/bin/activate  # On macOS/Linux
# or
.venv\Scripts\activate  # On Windows

# Install dependencies
uv sync

# Start the backend
./start-app.sh  # On macOS/Linux
# or
start-app.bat  # On Windows

The backend will be available at http://localhost:8080

Environment Variables (optional):

• OLLAMA_HOST: Ollama service URL (default: http://localhost:11434)
• OLLAMA_MODEL: Model to use (default: gemma3:12b-it-qat)
• ENVIRONMENT: Set to production or local (default: local)

2. Ollama Setup

Option A: Local Installation

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Start Ollama
ollama serve

# Pull Gemma model
ollama pull gemma3:12b-it-qat

Option B: Docker

cd ollama-gemma
docker build -t revela-ollama .
docker run -p 11434:11434 revela-ollama

3. Chrome Extension Setup

cd chrome-extension

# Install dependencies
npm install

# Build the extension
npm run build

Install Extension:

1. Download the extension: revela_chrome_extension.zip
2. Extract the zip file
3. Open Chrome and navigate to chrome://extensions/
4. Enable "Developer mode" (toggle in top right)
5. Click "Load unpacked"
6. Select the extracted extension folder

Production Deployment

Prerequisites

• Google Cloud Platform account
• gcloud CLI installed and configured
• Required GCP APIs enabled:
  • Cloud Run API
  • Cloud Build API

Setup GCP Project

# Set your project ID
export PROJECT_ID="your-project-id"
gcloud config set project $PROJECT_ID

# Enable required APIs
gcloud services enable run.googleapis.com
gcloud services enable cloudbuild.googleapis.com

Deploy Ollama Service (GPU Instance)

cd ollama-gemma

gcloud run deploy revela-ollama \
  --source . \
  --region europe-west4 \
  --platform managed \
  --no-allow-unauthenticated \
  --cpu 4 \
  --memory 8Gi \
  --gpu 1 \
  --gpu-type nvidia-l4 \
  --timeout 600 \
  --max-instances 5 \
  --min-instances 0 \
  --concurrency 4

# Get the service URL
export OLLAMA_HOST=$(gcloud run services describe revela-ollama \
  --region europe-west4 \
  --format='value(status.url)')

Note: GPU instances require quota allocation. Request NVIDIA L4 GPU quota under Cloud Run API if needed.

Deploy Backend Service (CPU Instance)

cd revela-app

gcloud run deploy revela-app \
  --source . \
  --region europe-west4 \
  --platform managed \
  --allow-unauthenticated \
  --cpu 2 \
  --memory 2Gi \
  --timeout 300 \
  --max-instances 10 \
  --min-instances 0 \
  --concurrency 80 \
  --set-env-vars OLLAMA_HOST=${OLLAMA_HOST}

# Get the backend URL
export BACKEND_URL=$(gcloud run services describe revela-app \
  --region europe-west4 \
  --format='value(status.url)')

Configure Service Authentication

Allow the backend to invoke the Ollama service:

# Get the backend service account
export BACKEND_SA=$(gcloud run services describe revela-app \
  --region europe-west4 \
  --format='value(spec.template.spec.serviceAccountName)')

# Grant invoke permission
gcloud run services add-iam-policy-binding revela-ollama \
  --region europe-west4 \
  --member="serviceAccount:${BACKEND_SA}" \
  --role="roles/run.invoker"

Deploy Chrome Extension

cd chrome-extension

# Build for production
npm run build

# The dist/ folder contains the production-ready extension
# Upload to Chrome Web Store or distribute as needed

License

This project is licensed under the MIT License - see the LICENSE file for details.