Open source · MIT licensed

Your documents.
Your machine.
Your answers.

A laptop-first AI assistant that finds answers across your files, cites its sources, and runs through a local LLM—without requiring a hosted AI API.

  • Local inference
  • Source citations
  • Configurable expertise
On-Prem RAG Assistant
Local
Supplier audit summaryGeneral Document Assistant
What are the required steps after a major nonconformity?
R

The procedure defines four required actions:

  1. Record and contain the nonconformity.
  2. Complete a root-cause analysis.
  3. Define corrective action and an owner.
  4. Verify effectiveness before closure.
1 CAPA Procedure · p. 4 2 Quality Manual · §10.2
Ask about your documents…
PDF
Knowledge base24 documents indexed
qwen2.5:7bRunning through Ollama
Built for private knowledge
PDFDOCXTXTMD
Indexed locally. Answered locally.
Capabilities

Document intelligence
without the cloud dependency.

Everything you need to turn local files into a searchable, conversational knowledge base.

01

Answers grounded in your files

Retrieval-augmented generation finds relevant passages before answering, with citations that point back to the source.

02

Local by design

Documents, embeddings, and chats stay on the configured host when Ollama runs locally.

03

Expertise you control

Create and save Agent Profiles that shape domain expertise, tone, and answer behavior.

04

Made for real documents

Index PDFs, Word documents, text, and Markdown—plus temporary attachments for a single chat.

05

Fast, streaming conversation

Read answers as they are generated and keep useful conversations in local SQLite history.

How it works

From document to cited answer.

A transparent local pipeline, designed to be inspected and extended.

01

Add documents

Choose files in the Knowledge Base or add an attachment to a chat.

02

Build the index

Text is split into searchable chunks and embedded into local ChromaDB.

03

Ask and verify

The local model answers from retrieved context and identifies supporting sources.

Your browserLocal applicationChromaDB+Ollama No hosted AI API required
Privacy posture

Keep sensitive context
inside your environment.

On-Prem RAG Assistant is built for a single trusted user running on a laptop or workstation. With a local Ollama endpoint, your content does not need to be sent to a hosted AI provider.

  • Local model inference and embeddings
  • Local document index and chat history
  • Inspectable, MIT-licensed source code

Local execution is a privacy control, not a compliance certification. Operators remain responsible for access control, encryption, retention, and regulatory requirements.

Quick start

Run it on your machine.

Python 3.10+, Ollama, and a few commands are all you need.

Read the full setup guide
Terminal
Windows scripts\setup.bat macOS / Linux chmod +x scripts/setup.sh && ./scripts/setup.sh Installs dependencies and downloads the default local models.
Open source

Build private AI
on your own terms.

Explore the code, run it locally, and help shape the roadmap.

Explore the repository