💬
LMIM Chat NEW
Encrypted peer-to-peer messaging between LMIM instances. Your keypair is your identity — no accounts, no servers reading anything.
- X25519 + NaCl crypto_box per message
- Blind relay — ciphertext only
- Groups via pairwise fan-out
- Browse the World: opt-in directory
🖥️
Bare Metal Terminal NEW
A real PTY-backed shell inside the dashboard. Your $SHELL, your config, your permissions — session re-attaches on refresh.
- Real PTY sessions, not a sandbox
- Persists across refresh / brief drop
- LMIM_TERMINAL prompt, aliases intact
- Idle sessions reaped after 4h
📱
Mobile Companion NEW
Native Android app that finds your instance via mDNS and mirrors the live dashboard over your own Wi-Fi.
- mDNS auto-discovery + manual fallback
- Backend-served — no APK rebuilds
- Full chat, groups, RAG, agenda
- IP caching with safe fallback
⚡
Speculative Decoding MTP
Optional draft-model acceleration. Small model drafts, main model verifies in batches — real GPU throughput gains, no correctness change.
- Draft-model dropdown in Settings
- Writes straight to .env
- Fully optional (leave on None)
- Bundled in XIPE_B build
🔨
Build Crew REBUILT
Type lmim build <description>. Planner → Builder → Inspector, rebuilt for true multi-file integration.
- Sequential multi-file builds
- Project-wide integration test
- Traceback-aware Inspector
- False-positive detection
⚡
Raw Mode NEW
Pure inference. One button strips prompts, memory, tools, and RAG. Nothing saved, ever.
- ~30-token identity prompt — that's it
- Completely ephemeral · zero history
- Direct llama-server call
- Red scanline · terminal prompt
🔒
Encrypter
AES-256-GCM for text, files, and whole directories. On-device, passphrase never stored.
- 100k PBKDF2 iterations · 256-bit keys
- .lmim container — documented format
- Browser + backend interoperable
- Batch directory mode · live progress
📄
Hybrid RAG v2
Drop a document, ask anything. Semantic + BM25 retrieval catches concepts and exact matches alike.
- Whole-document mode for short docs
- BM25 + embeddings fused via RRF
- MMR reranking · adaptive top_k
- Chunks respect headings + code
📁
Workspace Sandbox
Point LMIM at a folder — it reads, writes, creates, edits. Path traversal blocked at the backend.
- safe_path() — traversal rejected
- Native folder picker via Electron
- Live file tree in sidebar
- Ask LMIM about the project
🕷
Web Scraper
Batch 10 URLs in parallel, targeting article content and filtering the filler.
- Basic or LMIM-analysis mode
- Respects robots.txt · 30s timeout
- Export JSON · copy · push to chat
- Rate-limited per domain
🎙
Voice Control
Talk to LMIM, hear it back — fully on-device. No cloud, no API key, no audio leaving the machine.
- Whisper.cpp STT on GPU / CPU
- Piper TTS — 5 languages
- Device routing · setSinkId
- Auto-speak toggle
💬
5-Channel Messaging
WhatsApp, Telegram, Email, Slack, Discord — all from one Service Switch panel.
- WhatsApp · Baileys · QR pairing
- Telegram · Bot API
- Email · SMTP / IMAP
- Slack + Discord · webhooks
📇
Contacts
Say names, not numbers. "Send WhatsApp to Maria García" resolves automatically.
- Natural-language name resolution
- Local address book
- Wired into every messaging tool
- Works with Campaign Blaster
⚡
CUDA Lightning
Auto-detects your NVIDIA GPU. 3–15× faster than CPU, no config.
- llama.cpp + whisper.cpp on CUDA 12.x
- RTX 20xx · 30xx · 40xx
- Falls back to CPU cleanly
- CPU / GPU toggle in Settings
🐧
Cross-Platform
One file. Linux AppImage + Windows Installer, both live. Bundled model, CUDA included.
- Ubuntu · Fedora · Debian · Arch · Pop!_OS
- Windows 10 / 11
- CUDA on both platforms
- Same features · same $0