Guardrails Local RAG Bot

Ollama Connection

Checking…

Endpoint URL

Active Model

Safety & Privacy

Enable Guardrails

Sensitivity Level

INTERNAL

Suitable for internal business data. Blocks credential and API key exposure.

Sensitivity Reference

Chunking Parameters

Chunk Size (chars)

Chunk Overlap (chars)

Document Library

Previously indexed collections — load without re-uploading.

No indexed collections yet.

Document Upload

Drop files here or browse

PDF · TXT · DOCX · DOC

Chat with Your Documents

Upload any PDF, TXT or Word file — your data never leaves your machine. The AI reads only what you share.

Upload Your Document

Drop a PDF, TXT or DOCX. It's split into small chunks and converted into vector embeddings stored locally with FAISS.

Ask a Question

Your question is embedded and the most relevant chunks are retrieved from the vector store — this is the Retrieval in RAG.

Local LLM Answers

The retrieved context + your question go to Ollama running on your machine. No cloud, no data leaks, fully private.

Guardrails Filter

Both input and output pass through NeMo Guardrails — blocking jailbreaks, credential leaks and sensitive data exposure.

PDF · TXT · DOCX

Local LLM (Ollama)

NeMo Guardrails

FAISS Vector Cache

100% Private