Looks Like This Service Is On Hold
I will build an ai chatbot trained on your documentation with rag


About this gig
Your users keep asking the same questions your docs already answer.
Support tickets pile up. Sales leads get lost in contact forms. And your docs, which took you months to write, sit there unread.
I build AI chatbots that live on your site and answer questions using YOUR documentation as the source. Every answer cites the exact page it came from, so your users trust it and your team can audit it.
See it in action: https://ask-the-docs-juliancano.vercel.app
(Live demo trained on Supabase's full docs, 1,823 chunks indexed)
WHAT YOU GET:
- Chatbot trained on your docs (site, GitHub, PDFs, help center)
- Standalone chatbot deployed to your subdomain
- Streaming answers with inline citations to the source
- Lead capture when users show buying intent
- Rate limiting built-in, no surprise API bills
- Code shipped to your GitHub, you own everything on day 7
STACK: Next.js, TypeScript, Claude Haiku 4.5, OpenAI embeddings, Supabase pgvector, Upstash Redis
Senior developer. Same system I built in 9 days, delivered to you in 7.
Message me before ordering to confirm scope and fit. Let's build something your users actually use.
Get to know Julian Cano
FullStack Web Developer
- FromMexico
- Member sinceMay 2020
Languages
Spanish
My Portfolio
FAQ
Will the chatbot make up answers (hallucinate)?
No. Every answer is grounded in your actual documentation. The bot retrieves relevant chunks from your docs first, then generates the answer from those chunks, and cites the source page inline. If the answer isn't in your docs, the bot says so instead of inventing one. You can verify every response
What's the ongoing cost after you deliver?
You pay for three things: Vercel hosting (free tier works for most sites, around $20/mo for high traffic), OpenAI embeddings (one-time during ingest, usually $5-30), and Anthropic for the chat model (~$0.001 per user question with Claude Haiku 4.5). For a site with 1,000 queries per month, expect $5
What happens when my documentation changes?
Basic: you message me and I re-run the ingest (small fee). Standard: same, plus I document the process so your team can do it. Premium: I set up an automatic weekly re-indexing job so your bot is always current with zero manual work. In all tiers, the code is in your GitHub, so any developer on your
Can I switch to a different LLM later or self-host?
Yes. The code is yours on day 7, shipped to your GitHub, deployed to your Vercel. The LLM is one environment variable (currently Claude Haiku 4.5). You can swap to OpenAI, Gemini, or a self-hosted model by changing the config. No framework lock-in, no proprietary pieces. You own everything.
What kinds of documentation does this work with?
Any text-based source: public docs sites (auto-scraped), Markdown in a GitHub repo, PDFs, Notion exports, Confluence exports, help center articles (Intercom, Zendesk, HelpScout). If your docs are in a unusual format, message me with a sample before ordering and I'll confirm compatibility before you
Who owns the source code?
You do. The code is delivered to your GitHub on day 7. You can read it, modify it, extend it, or hand it off to your own developers. Use it internally for your product however you want. Redistribution or reselling the system as a service to third parties requires a written agreement, so please reach
What's included in the 30-day post-launch support (Premium)?
Included: bug fixes, performance tuning, re-indexing when your docs change, answering technical questions about the code, and one 1-hour session to adjust prompts and improve accuracy based on real user queries. Not included: new features (additional integrations, new UI components, extra languages

