ComingUp ComingUp
LLM Connect

LLM Connect

An offline AI runner, multi-provider client

Is this your project? Ask us to take it down

Gallery

LLM Connect

About

Local & Cloud AI

- Offline Models: Run quantized GGUF or LiteRT models fully offline.

- Multi-Provider Cloud: Use Ollama, OpenAI, Claude, Groq, Grok, or Gemini.

- Encrypted Keys: API keys are stored in encrypted on-device storage.

Agents (LLMC Bots)

- Autonomous Tasks: Give each agent a goal, system prompt, and follow-up steps. It runs in the background.

- Flexible Scheduling: Run once, on an interval, daily, or on chosen weekdays. Limit by cycle count or time, or run on demand.

- Per-Agent Models: Each agent uses its own cloud model or an on-device model.

- Run Logs: Every agent keeps a log of runs, messages, and errors.

Tools & Function Calling

- Built-In Tools: Files, device information, and time. Disabled by default, so you choose what each chat or agent can use.

- Device Manager: Lets the model query the packages installed on your device.

- Custom Tools: Build web-request tools from templates for search, dictionary, authenticated APIs, and webhooks.

LLMC Terminal

- Real Ubuntu 24.04: A full Linux shell with apt, Python, git, curl, and more.

- One-Tap Setup: The base system downloads and configures on first launch.

- Shared AI Files: Files from chats and agents appear in /root/ai-files.

- Touch Controls: A key bar with Ctrl, Alt, Esc, Tab, arrows, and paste. Pinch to zoom.

- Select & Act: Copy highlighted text, select all, or open links.

- System Logs: Search and filter Ubuntu logs, including apt and dpkg, plus your session.

- Quick Commands: A reference of common commands with one-tap copy.

- Reset Anytime: Wipe the environment. A confirmation step prevents accidents.

Local API Server

- Share Your Model: An OpenAI-compatible endpoint that devices on your Wi-Fi can call.

- Token Protection: Every request needs an access token. Rotate it anytime.

- Streaming: Standard chat completions with streaming. Most apps connect by changing the base URL.

- Background Operation: A foreground service keeps the server reachable.

RAG & MCP

- Retrieval-Augmented Generation: Ground answers in your own documents.

- Model Context Protocol: Connect MCP servers for external tools and data.

Speech Input

- On-Device Speech-to-Text: Dictate prompts with Whisper models that run on your phone.

Model Management

- Import & Guardrails: Import model files with RAM safeguards to protect stability.

- Hardware Acceleration: Choose CPU, GPU, or NPU.

- Multimodal: Process images with vision-language models and mmproj files.

Chat & Generation Controls

- Live Streaming: Pause or stop mid-stream and keep the output so far.

- Tuning: Adjust system prompt, context window, max tokens, temperature, seed, Top-P, and Top-K.

- Presets & Metadata: Code/Math, Creative, and General presets. Each reply shows token count, speed, seed, and duration.

Conversations & Statistics

- On-Device Storage: Chats live in a local SQLite database.

- Chat Management: Search, title, rename, and manage history privately.

- Chat Statistics: A collapsible panel with messages, tokens, speed, duration, server requests, tool calls, and errors.

Monitoring & Benchmarks

- Live Telemetry: RAM, available memory, and device temperature.

- Hardware Benchmarks: Token speed, temperature profiles, and performance scores.

Guidance

- System Intents: Open model files from third-party file managers.

- Guided Tour: Covers hardware, models, agents, tools, and the terminal.

Turn your Android device into a private AI workstation.

Comments (0)

No comments yet. Be the first to comment!