Open-source AI workspace · Nuxt 4 + Paperless-ngx

Private document
intelligence
for your AI workflows.

Taan Mind connects chat, OCR, metadata enrichment, and KPI dashboards around Paperless-ngx — with local-first options through Ollama and a Docker-ready stack.

Start locally View GitHub

OSS: Community project
OCR: Ollama + MuPDF
KPI: Document dashboards

Product surface

Chat, documents, OCR, and metrics in one workspace.

A technical companion for teams that want useful document AI without losing control of their files.

local-first ready

Privacy-focused

Designed around local storage and controlled AI providers.

Paperless companion

Uses Paperless-ngx as the durable document system.

Docker-ready

Ships with a full-stack compose workflow for evaluation.

Developer-friendly

Nuxt, Nitro APIs, typed data access, and clear boundaries.

Latest release

v1.0.15

Inline intelligence now lives where work begins.

The latest Taan Mind release brings contextual AI actions into the prompt flow, while preserving the same local-first document pipeline and explicit model registry.

Compose

Inline AI Assistant

Rewrite, refine, and trigger document-intelligence actions directly while drafting messages.

Context

Ordered document grounding

Build assistant context from explicit document selections so answers stay tied to the right Paperless records.

Hierarchy

Cleaner chat controls

Assistant actions sit above model, personality, and document selectors for a calmer workspace.

Designed for responsible document AI

Paperless-ngx → OCR → AI context → assistant actions → measured outcomes.

Server-side Paperless access

HTTP-only anonymous sessions

Hosted and local model choices

Docker-ready evaluation path

Product tour

Screens that explain the workflow.

Real product surfaces show the bridge between Paperless-ngx, AI chat, processing status, and KPI inspection.

How it works

A clear pipeline from archive to answers.

The product story is simple: synchronize documents, process them with OCR and AI, then use the enriched context in chat and dashboards.

Connect Paperless-ngx

Proxy document operations and sync metadata into the app cache.

Process with OCR + AI

Extract text, clean content, and suggest titles, tags, correspondents, and types.

Ask, inspect, and improve

Inject document context into chat, review KPIs, and patch metadata back to Paperless.

Core capabilities

Built as three focused layers.

Instead of a generic feature grid, Taan Mind is easier to understand as a stack: conversation, document intelligence, and operations.

09 Capabilities

Conversation layer

A document-aware AI chat surface with selectable providers and personalities.

Streaming AI chat

Use MiniMax, GLM, Nova, and available non-OCR Ollama models.

Document context

Inject OCR/AI-processed Paperless content directly into chats.

Inline AI actions

Rewrite prompts and trigger document-intelligence actions before sending.

Document intelligence

A processing pipeline that turns archived files into useful structured context.

Paperless-ngx proxy

CRUD, search, and binary download behind the app API.

Automatic OCR

Background extraction with Ollama OCR models and MuPDF page processing.

AI metadata extraction

Suggest titles, tags, correspondents, and document types from content.

Operations layer

Everything needed to evaluate, run, and maintain the stack responsibly.

KPI dashboard

Charts for status, timeline, MIME type, and document type distribution.

Anonymous sessions

HTTP-only session cookies with app-owned local SQLite storage.

Docker-ready runtime

Hardened multi-stage Dockerfile and integrated Paperless stack via Compose.

Implementation stack

Modern Vue infrastructure, practical AI plumbing.

The official app is built with Nuxt 4 and a pragmatic set of tools for APIs, storage, OCR, rendering, and dashboards.

↗ Nuxt 4 Full-stack Vue 3 framework ↗ Nuxt UI 4 Component library on Tailwind CSS 4 ↗ AI SDK Streaming AI integration ↗ Drizzle ORM Type-safe SQLite ORM ↗ NuxtHub Database and deployment foundation ↗ Ollama Local LLM runtime for OCR/chat models ↗ MuPDF PDF and image page processing ↗ Comark Markdown rendering ↗ nuxt-charts Chart.js visualizations

Built in the open

Open source & self-hostable

Run the whole stack on your own infrastructure — no vendor lock-in, no per-seat fees. Inspect the code and contribute on GitHub.

Explore the repository

Installation

Clone it.
Run it locally.

Keep the path familiar for open-source contributors: install dependencies, copy the environment file, or launch the full Paperless stack with Docker Compose.

Prerequisites

→ Node.js 22+
→ pnpm 10+
→ Docker & Docker Compose

LOCAL DEVELOPMENT

git clone https://github.com/zademy/taan-mind.git
cd taan-mind
pnpm install
cp .env.example .env

DOCKER COMPOSE (FULL STACK)

# Build and start everything
cp .env.example .env
docker compose up -d --build

# Optional: remove the completed bootstrap container after startup
docker compose rm -f paperless-bootstrap

Community support

Help sustain open-source development.

Taan Mind grows through community use, feedback, maintenance, and contributions. Donations help keep improvements moving without turning the project into a closed product.

Support on Ko-fi Support with Stripe