What it does WebWhiz crawls your public website, chunks the text, generates embeddings, and serves a ChatGPT-backed chatbot you embed with a single script tag. The self-hosted route needs Docker or a manual install of NestJS, two Celery workers, MongoDB, and Redis.

The interesting bit It skips the dedicated vector database entirely—embeddings live in Redis with cosine similarity handled by a Python worker. For “a few hundred or thousands of pages” the README claims this is fast enough; if you outgrow it, you email them.

Key highlights

• Monthly re-crawl of your site (more frequent requires asking)
• Two OpenAI keys supported: primary and failover
• Python worker handles HTML/PDF extraction and similarity math
• React frontend, NestJS backend, PM2 orchestration
• AGPLv3 license

Caveats

• The “bring your own OpenAI key” FAQ says “not at the moment, but in a couple of days”—unclear if this shipped
• No mention of actual token costs, rate limits, or embedding model choice in the README
• Redis-as-vector-store is a deliberate trade-off, not a feature for large corpora

Verdict Worth a spin if you want a quick, self-hosted support bot and your docs fit in Redis. Look elsewhere if you need real-time sync, multi-tenant scaling, or clarity on pricing mechanics.