I will build a rag ai chatbot for your pdfs and docs that cites every answer


About this gig
Your team keeps asking the same questions, and the answers are buried in a 60-page handbook, a folder of PDFs, or an old help center. A RAG chatbot reads those documents and answers in plain language, with the page each fact came from, so anyone can check it.
I build these end to end. I've shipped a live RAG assistant that answers only from its source texts, cites every answer, and switches between several models automatically if one fails.
What you get:
- Answers grounded in your PDFs, Word files, web pages, or help center
- A citation on every answer that opens the exact page
- An honest 'I couldn't find that' when the answer isn't there
- Testing on your own real questions before handover
- A monthly running cost estimate, billed to your own accounts
Models: OpenAI, Claude, Gemini, Llama, or Mistral.
Search: Pinecone, pgvector, Supabase, Qdrant, Weaviate, Chroma, or MongoDB Atlas.
App: LangChain or LlamaIndex, with Python, FastAPI, Node.js, or Next.js.
Not sure what fits? Send me one sample document and a few questions people ask. I'll tell you honestly what's possible.
Get to know Emon
Software Engineer
- FromBangladesh
- Member sinceJan 2025
- Avg. response time1 hour
Languages
Bengali, English
My Portfolio
FAQ
Will it make things up?
It's built to answer only from your documents. If nothing relevant turns up, it says so and shows the closest page instead of guessing. I also test it with questions your files can't answer to make sure it declines them.
Who pays for the AI and the database, and how much?
You do it directly in your own accounts. I don't add a markup. For a small team, it's usually a few dollars to a few tens of dollars a month. Before we start, I'll estimate it for your expected number of questions using current prices.
Is my data safe? Will it be used to train AI models?
Your files stay in your own database and cloud account. The big AI providers don't train on API data by default, and I'll confirm the policy for the one you choose. If nothing may leave your servers, an open model can run on your own machine.
Which AI model should I use?
It depends on your budget, language, and privacy needs. I'll recommend one, but OpenAI, Claude, Gemini, Llama, and Mistral all work. The code is set up so you can switch models later without rebuilding anything.
How many documents can it handle?
Basic covers up to 20 PDFs, enough to prove it works on your content. Standard and Premium handle thousands of pages. Scanned documents need text recognition first, so let me know if you have any.
Can I add or update documents myself later?
On Standard and Premium, yes. You get an upload page to add, replace, or remove files, and the chatbot picks up the changes. On Basic, I include a short guide for re-running the import yourself.
Can it go on my website, Slack or WhatsApp?
Yes. A website chat widget, Slack, Teams or WhatsApp is an extra on Basic and Standard, and one is included in Premium. It can also sit behind your own login or plug into your app through an API.
What languages does it support?
Documents and questions can be in most major languages, and the answer comes back in the language of the question. I've built one that answers in both English and your preferred language.
How do you use AI tools?
I write code myself and also use AI coding assistants such as Claude Code. Every change is reviewed line by line and tested before delivery, and I take full responsibility for it. Your files go to an AI tool only if you agree. Want no AI help at all? Just tell me before ordering.

