Mobile-Artificial-Intelligence/maid
Maid
Maid is a mobile large language model chat client and local runner. It provides a unified interface for interacting with AI models via cloud APIs or by executing model files directly on mobile hardware for offline generation.
The project functions as a multi-provider manager that handles API keys and a system for downloading and loading curated model files to the device. It utilizes a provider-agnostic interface to allow switching between local and remote backends.
The application includes tools for organizing and exporting chat histories, synchronizing user data across multiple devices, and tuning generation parameters such as temperature and context length.
Features
- Offline Chat Clients - Acts as a chat application that runs language models locally on the device for offline use.
- Local On-Device AI - Enables AI applications to run inference locally on the user's mobile hardware.
- AI Provider Integrations - Includes connectors for integrating external AI model providers using API keys.
- Cloud AI Integrations - Implements connectors for leveraging cloud-based artificial intelligence services via a mobile interface.
- Unified Provider Interfaces - Provides a unified interface bridging both cloud-based and locally hosted AI model providers.
- Local Model Execution - Provides the capability to run artificial intelligence models directly on local hardware for offline generation.
- On-Device LLM Runners - Functions as a dedicated runner that loads and executes LLMs on mobile hardware for text generation.
- Model Downloaders - Provides a system for browsing, downloading, and loading curated model files for on-device execution.
- On-Device Models - Uses frameworks and engines optimized for local execution of models on mobile hardware.
- Model Management - Provides a system for browsing, downloading, and configuring curated model files on the mobile device.
- Provider-Agnostic Model Interfaces - Provides abstraction layers that standardize inputs and outputs across multiple different LLM providers.
- Chat History Exporters - Includes utilities for saving and exporting conversation logs into structured or plain text formats.
- Generation Temperature Controls - Allows adjustment of the randomness of model output using a temperature parameter.
- Remote Inference Providers - Implements interfaces for offloading generative tasks to external, hosted model endpoints via network requests.
- Model Generation Tuning - Allows adjustment of LLM inference parameters such as temperature and token limits.
- Model Parameter Tuning - Provides controls for adjusting the deterministic or creative behavior of AI models via parameters.
- Cross-Device Synchronization - Mirrors user data and chat state across different mobile hardware devices.
- Cloud File Sync - Synchronizes local user profiles and chat logs with remote cloud storage for cross-device consistency.
- User Profile Synchronization - Synchronizes user-specific preferences and interaction history across different mobile devices.
- Self Hosted Platforms - Self-hosted AI chat application.
