GPU Listings Backend

A simple Flask backend for fetching and managing GPU listings from various providers.

Quick Setup

1. Create a virtual environment and activate it:

python -m venv .startup
source .startup/bin/activate

2. Install dependencies from requirements.txt:

pip install -r requirements.txt

3. Install PostgreSQL if not already installed

sudo pacman -S postgresql

Note: These are for linux, if on mac, pacman will not work. You will have to use brew. Find out how to set up on Windows.

4. Create a .env file in the root directory:

DATABASE_URI=postgresql://postgres:postgres@localhost:5432/neotix

5. Now run the script 'setup_postgres.py' in the scripts folder.

Note: We have not yet tested this on a Windows machine. So, if you are on a Windows computer, read the file carefully then figure out how to create the database manually.

5. Initialize the database:

flask db upgrade

6. Run the application:

flask run

In parallel, run the command:

7. Fetch GPU data:

flask fetch-gpu-data

If it does not work on the get go. Run,

git clone https://github.com/Neotix-Dev/gpuhunt.git

As the gpuhunt repo is not automatically downloaded. Re-run the command flask fetch-gpu-data again.

API Documentation

Authentication

Most endpoints require Firebase authentication token in the Authorization header.

GPU Endpoints (/api/gpu)us

• GET /get_all - Get all GPU listingsUS national
• GET /get_gpus/<page_number> - Get paginated list of GPUs, start from page 1. Contains more info on the GPUs, better to use this endpoint as it pulls 200 GPUs at a time
• GET /<id> - Get specific GPU by ID
• GET /search?q=<query> - Search GPUs by name/specs with fuzzy matching
• GET /filtered - Get filtered GPUs with pagination
  • Query params: gpu_name, gpu_vendor, min/max_gpu_count, min/max_gpu_memory, min/max_cpu, min/max_memory, min/max_price, min/max_gpu_score, provider, sort_by, sort_order, page, per_page
• GET /<gpu_id>/price-points - Get current price points for a GPU
• GET /<gpu_id>/price-history - Get price history with optional date range
• GET /vendors - List all GPU vendors
• GET /hosts - List all GPU providers

User Management (/api/user)

• POST /register - Register new user with Firebase
  • Required: email, password, first_name, last_name, organization, experience_level
• POST /sync - Sync Firebase user with backend
• GET /profile - Get user profile and preferences
• PUT /profile - Update user profile
• POST /update - Update user data
• DELETE / - Delete user account

Projects (/api/projects)

• GET / - Get user's projects
• POST / - Create new project (requires: name)
• PUT /<project_id> - Update project details
• DELETE /<project_id> - Delete project
• POST /<project_id>/gpus - Add GPU to project
• DELETE /<project_id>/gpus/<gpu_id> - Remove GPU from project

User Preferences (/api/preferences)

Selected GPUs

• GET /selected-gpus - Get selected GPUs
• POST /selected-gpus - Add GPU to selection
• DELETE /selected-gpus/<gpu_id> - Remove GPU from selection

Favorite GPUs

• GET /favorite-gpus - Get favorite GPUs
• POST /favorite-gpus - Add GPU to favorites
• DELETE /favorite-gpus/<gpu_id> - Remove from favorites

Rented GPUs

• GET /rented-gpus - Get active rentals
• POST /rented-gpus - Add rented GPU
• DELETE /rented-gpus/<gpu_id> - End GPU rental

Price Alerts

• GET /price-alerts - Get price alerts
• POST /price-alerts - Create price alert
  • Required: gpuId/gpuType, targetPrice, isTypeAlert
• DELETE /price-alerts/<alert_id> - Remove price alert

A/B Testing (/api/analytics)

The backend implements A/B testing functionality to track and analyze user preferences between different view types (grid vs. table view).

View Preference Tracking

• POST /view-preference - Record user view preference
  • Required payload:

    {
      "sessionId": "unique_session_id",
      "timestamp": "ISO timestamp",
      "viewType": "grid|table",
      "initialView": "grid|table"
    }

Analytics

• GET /view-preference/summary - Get summary of view preferences
  • Returns:

    {
      "total_sessions": "number of unique sessions",
      "view_preferences": {
        "grid": "number of users who preferred grid view",
        "table": "number of users who preferred table view"
      },
      "conversion_rates": {
        "grid": "percentage who stayed with grid view",
        "table": "percentage who stayed with table view"
      }
    }

Implementation Details

• Users are randomly assigned either grid or table view on their first visit
• The system tracks:
  • Initial view type assigned
  • View type changes during the session
  • Final view type preference
• Analytics are stored in JSON format at data/analytics/view_preferences.json

Setup Instructions

1. Ensure the data/analytics directory exists:

   mkdir -p data/analytics

2. Initialize the analytics file:

   echo "[]" > data/analytics/view_preferences.json

3. Ensure write permissions:

   chmod 644 data/analytics/view_preferences.json

All the analytics data is stored in data/analytics/view_preferences.json.

Response Format

All endpoints return JSON with standard structure:

• Success: { data: [result] }
• Error: { error: "error message" }

Status Codes

• 200: Success
• 201: Created
• 400: Bad Request
• 401: Unauthorized
• 404: Not Found
• 500: Server Error

Test change