WebWhiz enables users to create AI chatbots trained on their website data, offering instant customer query responses without coding. It targets website owners and businesses seeking to enhance customer support by leveraging their existing content.

How It Works

WebWhiz automatically crawls a specified website URL to fetch and prepare training data, including text and metadata. It then trains a ChatGPT model on this data, creating a chatbot that can be embedded into a website via a script tag. The system utilizes a multi-component architecture: a NestJS API server, Python Celery workers for crawling and embeddings, and a JavaScript worker for content extraction. MongoDB and Redis are used for database and caching, respectively.

Quick Start & Requirements

• Self-Hosting (Docker):
  • Clone the repository.
  • Edit .env.docker with OPENAI_KEY and OPENAI_KEY_2.
  • Run docker-compose up (or docker-compose up -d for daemon).
  • UI available at http://localhost:3030, backend at http://localhost:3000.
• Manual Setup:
  • Prerequisites: MongoDB v6, Redis v7, Node.js v18+, Yarn, Python v3.6+.
  • Configure .env files for the API server and workers.
  • Install dependencies: yarn install (root), cd workers && python3 -m venv venv && source venv/bin/activate && pip install -r requirements.txt.
  • Run with PM2: yarn run build && npm install -g pm2 && pm2 start ecosystem.config.js.
  • Frontend setup: cd frontend, create .env, npm install, npm run start.
• SDK: Install via npm install webwhiz or use CDN: https://www.unpkg.com/webwhiz@1.0.0/dist/sdk.js.

Highlighted Details

• No-code builder for chatbot creation.
• Supports regular data updates (monthly crawl frequency).
• Custom data training via pasting content.
• Offline message handling.

Maintenance & Community

• Contact: hi@webwhiz.ai.

Licensing & Compatibility

• License: GNU Affero General Public License v3.0 (AGPLv3).
• SDK available on NPM, CDNs, and GitHub.

Limitations & Caveats

The project requires users to provide their own OpenAI API keys. Website crawling is currently limited to a monthly frequency, with more frequent scans requiring direct contact. Token limits per plan can result in chatbots providing predefined messages if exceeded.