I will build rag ai chatbot trained on your documents ai assistant


About this gig
I build RAG chatbots that answer from your own documents and cite the page they came from
Most "AI chatbots" are a prompt wrapped around ChatGPT they improvise when they don't know. Mine retrieve first: your content is chunked, embedded and stored in a vector database, so every answer traces back to your material
What you get: Ingestion for PDF, DOCX, web pages, Notion or a database Vector store you own (pgvector, Qdrant or Pinecone) Retrieval plus reranking, tuned on your real questions Chat widget for your site, or a clean REST endpoint Source citations and an honest "not in my documents" fallback An evaluation set, so accuracy is a number you can check Full repository, deployment and a handover call
Stack: Python, TypeScript, LangChain or LlamaIndex, OpenAI / Anthropic / open models, Next.js, Docker
You keep the code, the keys and the database. No lock-in, no monthly fee to me. Model API costs are billed to your own account, and I'll estimate them before you order.
Message me with your document types and roughly how many and I'll tell you what's realistic first
I am committed to delivering modern, reliable, and intelligent digital solutions that help your business growth
Get to know Nabil Ess
AI and Full Stack Developer
- FromMorocco
- Member sinceMar 2026
- Avg. response time1 hour
Languages
English, French
My Portfolio
Other AI Development Services I Offer
FAQ
Will it make things up?
It retrieves before it answers, and every response links to the source passage. When your documents don't cover a question, it says so instead of guessing. I also deliver an evaluation set so you can measure how often it's right.
Who pays for the OpenAI or Claude API?
You do, on your own account, so you keep control and pay wholesale. Typical cost is a few dollars per thousand questions on a small corpus. I estimate it for your specific volume before you order.
Can it run without sending data to OpenAI?
Yes. I can build it fully self-hosted with open models and a local vector store, which is the usual choice for legal, medical or internal HR documents. Choose the Platform package and mention it in your brief.
What document formats do you support?
PDF (including scanned, with OCR), DOCX, TXT, Markdown, HTML, CSV, Notion exports, and content pulled from a website or an API. Tell me what you have and I'll confirm before you order.
Do I get the source code?
Yes, in every package. You get the repository, the environment configuration and setup documentation. Nothing runs on my servers and there's no dependency on me after handover.
How many documents can it handle?
Prototype covers 50, Production covers 1,000, Platform is unlimited with a custom quote. Volume matters less than variety — 5,000 clean similar reports are easier than 200 mixed messy ones.
Can you embed it in my existing site?
Yes. You get either a script-tag widget that drops into any site, or a REST endpoint your developers call. WordPress, Webflow, Shopify and custom React apps are all fine.
What do you need from me to start?
A representative sample of your documents, 10–20 real questions you want answered, and where it should live. That's enough to begin.
