Lamp AI

Your personal AI assistant that remembers everything.

Desktop-first • Local storage • Persistent memory • Telegram integration

Download

---

What is Lamp?

Lamp is a native desktop AI assistant built on Electron. Unlike browser-based chatbots, Lamp runs locally, stores your data on your machine, and builds a persistent memory graph of facts and entities across all your conversations.

Think of it as a personal AI that actually knows you — the more you use it, the smarter it gets.

Core Ideas

Memory that persists. Lamp extracts facts from your conversations and maintains an entity graph. When you mention something weeks later, it already has context. No copy-pasting, no "as I mentioned before."

Threads, not just chats. Fork any message into a side thread to explore a tangent without losing the main conversation. Quote a specific part to branch from exactly the right point.

Integrations that talk. Connect Telegram and let Lamp read, search, and send messages on your behalf. Gmail, Calendar, and Apple Health are on the roadmap.

Your data, your machine. Everything lives in a local JSON store. No cloud sync, no telemetry, no surprises.

Features

Streaming chat	Real-time token streaming with thinking indicators
Threaded conversations	Main thread + side threads with quote-based branching
Persistent memory	Fact extraction, entity graph, relevance scoring
Web search	Perplexity Sonar Pro via OpenRouter
File attachments	Images, PDFs, code files — drag and drop
Telegram	List chats, read/send/search messages via MTProto
Command palette	⌘K to navigate, search chats, quick actions
Auto-updates	Seamless updates via GitHub Releases
Proxy support	Optional HTTP proxy for restricted networks

Architecture

┌─────────────────────────────────────────────────────┐
│                    Electron Shell                    │
├──────────────────────┬──────────────────────────────┤
│    Main Process      │       Renderer (React)       │
│                      │                              │
│  ┌────────────────┐  │  ┌────────────────────────┐  │
│  │ Agent Runtime   │  │  │ Chat UI                │  │
│  │  ├─ Supervisor  │  │  │  ├─ Sidebar            │  │
│  │  ├─ Tools       │◄─┼──┤  ├─ Threads            │  │
│  │  └─ Memory      │  │  │  ├─ Command Palette    │  │
│  ├────────────────┤  │  │  └─ Settings            │  │
│  │ Storage Layer   │  │  └────────────────────────┘  │
│  │  ├─ Orama (vec) │  │                              │
│  │  └─ JSON Store  │  │                              │
│  ├────────────────┤  │                              │
│  │ Integrations    │  │                              │
│  │  ├─ Telegram    │  │                              │
│  │  └─ OpenRouter  │  │                              │
│  └────────────────┘  │                              │
└──────────────────────┴──────────────────────────────┘

Multi-Agent System

Lamp runs a Supervisor Router that orchestrates multiple agents:

• Main Assistant — handles user queries, uses tools, streams responses
• Memory Curator — extracts facts, merges entities, archives stale knowledge

Each agent has a scoped tool policy. The main assistant can invoke the memory curator, but not vice versa. Tool calls are executed as async generators with progress events piped to the UI in real time.

Agent Tools

Tool	Description
web_search	Search the web via Perplexity Sonar Pro
search_messages	Semantic search across chat history
memory_query	Query the fact/entity graph
memory_upsert_fact	Create or update a fact
memory_merge_entities	Deduplicate entities
memory_archive_facts	Archive outdated facts
memory_link_identity	Link identities across channels
telegram_*	Full Telegram integration (5 tools)
invoke_agent	Call another agent from within a run

Tech Stack

Layer	Tech
Shell	Electron 33
Build	electron-vite, electron-builder
UI	React 18, TypeScript
Styling	Tailwind CSS 4, Radix UI
Icons	Lucide, Hugeicons
LLM	OpenRouter (chat, embeddings, vision)
Search	Orama (vector + full-text)
Telegram	MTProto via telegram package
Markdown	Streamdown
Updates	electron-updater

Getting Started

Prerequisites

• Node.js 22+
• OpenRouter API key (get one here)

Development

npm install
npm run dev

Build

# macOS
npm run dist:mac

# Windows
npm run dist:win

Built artifacts land in the dist/ folder — DMG + ZIP for macOS, NSIS installer for Windows.

Keyboard Shortcuts

Shortcut	Action
⌘ N	New chat
⌘ K	Command palette
⌘ ,	Settings
⌘ I	Integrations
⌘ ⌫	Delete current chat
⌘ .	Stop streaming

Roadmap

• Gmail integration
• Google Calendar integration
• Apple Health integration
• More LLM providers (direct OpenAI, Anthropic)
• Knowledge base with document ingestion
• Voice input

License

Private. All rights reserved.