🔒 All Phases · Privacy & Security

Local-First Research Workstation

Your Data Sovereignty. Your Academic Privacy.

InkCop's SQLite + ObjectBox + local vector database multi-DB architecture stores all unpublished papers and private knowledge bases entirely on your device. We never collect your academic work. Local LLM support enables offline-capable operation — in extreme cases, fully offline when you handle PDF/image text extraction manually.

Local Multi-Database Architecture

SQLite / SQLCipher
Structured Data

Conversation history, settings, citation records. AES-256 encryption supported.

ObjectBox + HNSW
Knowledge Base Vector Engine

Document chunk storage and millisecond nearest-neighbor search (local RAG).

Kuzu Graph DB
Knowledge Graph

Entity relationship network with multi-hop queries and cross-paper reasoning.

llama.cpp
Local LLM

Run Qwen / Llama and other open-source models completely offline.

What Needs Network? What's Fully Local?

Requires Network

  • Calling cloud LLMs (OpenAI / Claude / Gemini etc.)
  • PDF/Word to text conversion when using cloud OCR
  • PubMed / arXiv literature search and download
  • Citation verification (tracing PubMed/CrossRef etc.)

Fully Local, No Network

  • Knowledge base semantic search (vector + graph)
  • Document editing, Track Changes, snapshots
  • Local LLM inference (Llama / Qwen etc., after download)
  • All data storage: documents, chat history, citations
  • Workflow execution (without external API nodes)

Core Features

Designed for data security and high-performance local inference

Local Vector DB (ObjectBox+HNSW)

ObjectBox high-performance object database with HNSW vector indexing delivers millisecond semantic search. All data stored locally.

🤖

Local LLM Support

Built-in llama.cpp service to run Qwen, Llama and other open-source models locally. Zero network for fully offline operation.

🔐

Encrypted Workspaces

Each workspace maps to an isolated local directory. Supports SQLCipher AES-256 encryption and encrypted archive export.

📁

Multi-Workspace Management

Multiple independent workspaces for different research projects, with complete data isolation between projects.

🔄

Workflow Engine

n8n/Dify-like visual workflow builder. Automate repetitive processes — format conversion, batch summarization, periodic archiving.

🔌

MCP Tool Extension

Load third-party MCP tools to extend with external professional capabilities. Can be extended into a team collaboration backend.