I will build a custom ai chatbot for your documents using rag and fastapi


About this gig
Most basic RAG chatbots fail in production because simple vector search misses exact product codes, clauses, or follow-up context. I build production-ready FastAPI + Qdrant retrieval pipelines engineered for real data.
What I Build For You:
- Hybrid Search (Dense + BM25): Captures both conceptual meaning and exact part numbers, SKUs, names, and clauses that standard bots miss.
- Two-Stage Reranking: Every chunk is re-scored via a Cross-Encoder reranker before reaching the LLM, cutting hallucinations and returning exact source citations.
- Follow-Up Memory: Automatically rewrites multi-turn follow-up questions into standalone search queries.
- 100% Flexible LLM Backend: Works with OpenAI, Groq, OpenRouter, or private local models (Ollama, LM Studio, vLLM) so you control your data and API costs.
What You Receive:
- Async FastAPI + Qdrant microservice containerized with Docker Compose.
- Live endpoints to upload, index, and delete PDF/TXT files.
- Built-in HTML Demo Chat UI + Swagger docs to visually test uploads and citations (custom frontend design not included).
- Setup guide & 2 adjustments within scope.
Send me a message with a sample document to get started!
Get to know Mohamed Hany
Ai Engineer
- FromEgypt
- Member sinceMay 2021
Languages
English, Arabic
FAQ
Do I need to provide my own AI API key?
Yes. To ensure you maintain 100% control over your data and billing, you will need to provide your own OpenAI, Groq API key or I can configure it to run 100% locally via Ollama or LM Studio with zero API key costs.
What file formats are supported out of the box?
The base package supports standard, text-based PDF and TXT documents. If your project involves DOCX, CSV, Excel sheets, or scanned image PDFs requiring OCR, you can select the Extra File Format add-on at checkout or message me first for a custom quote.
How is the chatbot delivered and how do I run it?
You receive the complete source code containerized with Docker Compose, a built-in HTML testing UI, and a step-by-step setup guide. You can run the entire stack locally or on any cloud server (AWS, DigitalOcean, Hetzner) with a single command.

