English | 简体中文
Problems It Solves • Core Features • Quick Start
AI-Native Immersive Knowledge Workspace
A closed-loop system that seamlessly connects multi-source data ingestion, intelligent retrieval orchestration, and knowledge processing & accumulation. Build a dedicated digital brain on your local machine — one that responds instantly and thinks deeply.
 |
 |
|---|
Knowledge scatters across cloud platforms and local drives. Piecing together a project's full picture forces constant app-switching, severely breaking your flow state.
PDFs, Word docs, PPTs, and historical files pile up on your machine. These "dead data" lack an activation mechanism — never surfaced when you actually need them.
"Reading materials," "querying AI," and "writing new content" live in separate apps. New content can't seamlessly complement your existing knowledge base.
Complex questions spanning multiple documents need depth — not surface-level stitching. You need multi-round deep investigation that fills implicit context automatically.
LLM as core orchestrator. Breaks free from single-retrieval pipelines — autonomously selects tools, reads full documents, performs multi-round reasoning with supplementary retrieval. Delivers panoramic, well-substantiated research conclusions.
Vectorization, on-device embedding inference, and semantic indexing all run locally. Files reside in plaintext on local storage. Physical-level isolation for sensitive data — zero cloud privacy leakage.
All data stored as standard Markdown with Frontmatter metadata. Long-term readability, cross-system portability, no vendor dependency.
Preset APIs ready to go. Bring Your Own Key supported. Switch between model providers in real time based on task workloads.
Native integration with Notion, Feishu, Apple Notes, and 6+ external knowledge domains. Multi-modal high-fidelity parsing (DDU) with silent incremental sync. Breaks down information silos across SaaS platforms.
One-click persistence of conversation insights into local Markdown docs. Ephemeral interactions → sustainable knowledge production. New content immediately feeds back into your semantic network.
Real-time streaming of reasoning logic and Function Call chains. Every claim hard-linked to precise source jump links. Strips the AI black box — full verifiability.
Expose as a standard MCP Server via HTTP streaming. Extends Luminite from standalone app into a system-level context module — external tools can invoke your local knowledge network directly.
Semantic search + keyword full-text search + hybrid retrieval, combined with entity recognition at ingestion. Discovers non-intuitive associative structures between isolated concepts.
Dynamic branch management, context truncation, node tracing, partial re-rendering. Image attachments and long-text injection. Precise fine-tuning control over the LLM context window.
FSM-based filesystem debounce engine + multi-stage pipelines + periodic orphan reconciliation. Strong consistency across retrieval indices, local DB, and file layers under high-frequency concurrent R/W.
Customized extraction per data source (e.g., Notion block-level API adaptation). Six-level recursive semantic slicing preserves document context coherence and section topology.
- Visit the official website to download the latest version for macOS.
- Drag
Luminite.appinto your Applications folder.
- Launch Luminite and complete the onboarding guide.
- Configure your LLM provider (use preset APIs or bring your own key).
- Connect data sources — local folders, Notion, Feishu, Apple Notes, etc.
- Ask questions: Open the conversation panel and query across all your connected knowledge.
- Ingest documents: Drop PDFs, Markdown, Word, or PPT files into Luminite — they'll be indexed automatically.
- Save insights: One-click persist any conversation insight back into your local knowledge base.
Enable the built-in MCP Server to let external AI tools (e.g., Claude, Cursor) access your local knowledge network directly.