1. Builds
  2. AI chatbot app over your own content

AI chatbot app over your own content

A deployed chat assistant that answers from your own docs, FAQ and product pages. It stores each user's history, runs moderation on input and has a spend cap on the AI API.

Priced from vendor pages
Prices checked
16 Oct 2026
Updated
17 Oct 2026
Exchange rate
1 USD = £0.7489
Sources
7
How we check

The stack

HOSTVercelNext.js native, Secret env vars£14.98
Layer
Hosting
Plan
Pro $20/seat/mo
What it does here
Next.js native, Secret env vars
Note
Hobby is non-commercial only, so it isn't used here
HOSTNetlifyCheaper hosting optionFree / £6.74
Layer
Hosting (alt)
Plan
Free $0 (300 credits) or Personal $9
What it does here
Cheaper hosting option
Note
Whether Netlify Free allows commercial use is not confirmed
DATAAUTHSupabasePostgres, auth and RLS in oneFree / £18.72
Layer
DB + auth + history
Plan
Free (50,000 MAU, 500 MB DB) or Pro from $25
What it does here
Postgres, auth and RLS in one
Note
Free pauses after 1 week of inactivity
AIOpenAIgpt-5-mini for answers, free moderationusage
Layer
AI API
Plan
Usage only: gpt-5-mini $0.25 in / $2.00 out per 1M tokens; gpt-5-nano $0.05 / $0.40; omni-moderation free
What it does here
gpt-5-mini for answers, free moderation
Note
No free tier found. VAT treatment not recorded
——Finds relevant content chunksNot yet priced
Layer
Retrieval / embeddings
Plan
—
What it does here
Finds relevant content chunks
Note
Excluded from totals until priced
OPSGitHubDependabot + secret scanningFree
Layer
Monitoring and security
Plan
Free
What it does here
Dependabot + secret scanning
Note
Secret scanning + push protection free on public repos; private Secret Protection price not found

Steps

  1. Scaffold a Next.js chat UI and deploy it on Vercel Pro (or Netlify).

    Done when the empty chat page loads at your live URL.

  2. Create Supabase tables for users, conversations and messages, with RLS so each user sees only their own rows (docs).

    Done when a second test user can't read the first user's messages.

  3. Ingest your content and retrieve the relevant chunks for each question. This retrieval approach is not priced.

    Done when a question about your docs returns the right passage.

  4. Call gpt-5-mini from the server only, and keep the API key as a Secret env var (docs).

    Done when the bot answers, and the key doesn't appear in the browser bundle or network tab.

  5. Run user input through omni-moderation (free) before the model call.

    Done when a flagged test input is blocked before it reaches the model.

  6. Set an OpenAI project budget or limit, and log tokens per conversation.

    Done when the budget is set and token counts show up per conversation.

Watch out for

  • Vercel Hobby is non-commercial, so plan on Pro from your first paid or business user.
  • Token cost grows with history length, so trim or summarise history.
  • Supabase Free pauses after a week of inactivity, so a live bot needs Pro.
  • Never expose the OpenAI key in client-side code.

What it costs each month

Stage Platform (no Lovable) Stripe fees Total + Lovable Pro
(a) Free (Netlify Free + Supabase Free) £0.00 £0.00 £0.00 £22.00
(b) Paid, lean (Netlify Personal £6.74 + Supabase Pro £18.72) £25.46 £0.00 £25.46 £47.46
(c) Paid, recommended (Vercel Pro £14.98 + Supabase Pro £18.72) £33.70 £0.00 £33.70 £55.70
(c) + assumed tokens (adds £4.87 of OpenAI usage, assumption) £38.57 £0.00 £38.57 £60.57
  • Token usage is an assumption, not a measurement. The assumed volume is 10M input + 2M output tokens a month on gpt-5-mini: $2.50 + $4.00 = $6.50, which is £4.87.
  • Excluded: VAT (USD vendors quote ex-VAT), usage overages, domains, retrieval/embeddings (not yet priced), and Stripe fees if you charge for the bot.
  • Neither total uses Vercel Hobby, because a monetised or business chatbot counts as commercial use.

Still open

  • The token volume (10M in / 2M out) is an assumption. Real usage is unknown until a hands-on test.
  • Retrieval/embeddings cost and any vector add-on are not yet priced.
  • Whether Netlify Free allows commercial use is not confirmed on Netlify's pricing page.
  • OpenAI VAT treatment is not recorded. Nothing has been tested hands-on.

Sources

Change log

Date Change Source
2026-10-16 New desk-costed build Stackguide desk research
2026-10-17 FX label clarified as 15 Oct 2026 close Editorial update
2026-10-23 USD prices re-converted at the Bank of England reference rate, 1 USD = 0.7489 GBP (16 Oct 2026), the one rate used site-wide (was 0.7535, exchangerates.org.uk 15 Oct 2026 close); £ figures and totals re-derived Bank of England series XUDLUSS, checked 2026-10-23