RAG Document Agent Development by Dmitrii KrotovRAG Document Agent Development by Dmitrii Krotov

RAG Document Agent Development

Dmitrii Krotov

AI Agent Engineer

ML Engineer

AI Developer

FastAPI

Google Gemini

Streamlit

Government & Public Administration

RAG Document Agent

A local Retrieval-Augmented Generation (RAG) system for answering questions strictly from provided documents.

Ingests real-world documents (technical, legal, literary)

Chunks and embeds them using Sentence Transformers

Stores embeddings in FAISS

Retrieves relevant passages with cosine similarity + threshold gating

Generates answers using Gemini only when evidence exists

Embeddings: sentence-transformers/all-MiniLM-L6-v2

Vector store: FAISS (IndexFlatIP)

LLM: Gemini 2.5 Flash

API: FastAPI

UI: Streamlit

No-answer refusal when no relevant context is found

Minimum similarity score enforced before LLM call

Context-only generation (no free hallucination)

pip install -r requirements.txt
uvicorn api.main:app --reload
streamlit run app/streamlit_app.py

Technical specifications (RFCs, APIs)

Legal / license documents

Long-form public-domain texts

Like this project

Posted Mar 16, 2026

Developed a RAG system for accurate document query answers.

Likes

Views