andrewyng/aisuite
Aisuite
This project is a framework for managing generative AI services through a unified provider interface and adapter layer. It provides a standardized API for calling multiple cloud-based and locally hosted models, translating provider-specific parameters and responses into a uniform format.
The system includes an agent orchestrator designed for long-running tasks, featuring state persistence for resuming runs and execution tracing to monitor decision-making processes. It integrates the Model Context Protocol to connect models to external servers and filesystems and employs a policy-based execution system with approval lists to control tool calling.
Additional capabilities cover automated tool execution through schema generation, local desktop automation, and speech-to-text transcription. The project also provides a conversational coding interface for file editing and shell command execution, as well as specialized subagents for read-only code review.
Features
- AI Agent Orchestration Frameworks - Provides a framework for managing long-running agent tasks with state persistence and execution tracing.
- AI Provider Interfaces - Offers a standardized interface for integrating both cloud-based and locally-hosted generative AI services.
- Agent State Persistence - Coordinates long-running AI tasks by persisting execution history and artifacts to allow resuming from specific points.
- AI Agent Orchestrators - Coordinates complex, long-running AI agent tasks using sandboxed toolkits, state persistence, and execution tracing.
- AI Provider Integrations - Connects to diverse generative AI services through a single interface to eliminate provider-specific code.
- Model Context Protocol - Integrates the Model Context Protocol to connect AI models to external filesystems and third-party servers.
- Model Request Routing - Implements a routing mechanism to direct prompts to different AI backends using a consistent parameter set.
- AI Provider Adapters - Provides a standardized adapter layer for integrating various generative AI model providers.
- AI Tool Execution - Enables AI models to trigger external functions and access local files using standardized schemas and policies.
- Chat Completion Services - Provides a standardized API for managing chat completions across various cloud and local AI services.
- Local Model Integrations - Provides integration layers for interfacing with locally hosted models to perform inference without external API keys.
- Model Provider Adapters - Standardizes diverse AI service APIs into a uniform request and response format using a common adapter structure.
- Provider-Agnostic Model Interfaces - Ships a standardized API layer for calling multiple generative AI services and local models.
- Agent Execution Tracing - Generates detailed trace output for agent activities to monitor the reasoning process in a visual viewer.
- Execution Control Policies - Filters model tool calls through a configurable allow-list and approval system to control host code execution.
- Local AI Inference - Provides an interface for executing machine learning models directly on local hardware resources.
- Model Context Protocol Clients - Includes a client for connecting large language models to external servers and filesystems via the Model Context Protocol.
- Streaming Response Aggregators - Aggregates real-time incremental text outputs from multiple AI providers into a unified data stream.
- Tool-Calling Schemas - Translates function signatures and JSON specifications into schemas compatible with AI model tool calling.
- Local Model Deployment - Supports interfacing with locally hosted generative AI models to perform inference without cloud API keys.
- Function Schema Generators - Automatically generates tool schemas from function signatures to enable AI models to execute functions.
- Agent Command Line Interfaces - Provides a command-line interface for interacting with a coding agent to edit files and execute shell commands.
- Local Desktop Agents - Executes local agents with filesystem access and tool connectors to produce documents and artifacts on a machine.
- Action Approval Policies - Employs approval policies and allow lists to control which tools a model is permitted to execute.
- Agent Execution Tracing - Records sequential logs of agent decisions and tool outputs for visual playback and debugging of reasoning processes.
- Unified AI Stream Aggregators - Aggregates incremental text outputs from different AI providers into a consistent real-time data stream.
- Language Model Development - Unified interface for interacting with multiple generative AI providers.
- LLM Development Utilities - Unified interface for multiple generative AI providers.
