DOCSEARCHbyAlgolia
Submit Docs

AIGNE Framework

repository·main·Indexed 17 days ago

https://github.com/aigne-io/aigne-framework

A functional AI application development framework for building modern agentic applications. It features the Agentic File System (AFS), a virtual file system that provides a unified path-based API for managing data across various backends. The framework includes modular components such as @aigne/afs-history for conversation tracking, @aigne/afs-git for Git repository integration with read-only and read-write modes, and @aigne/afs-explorer for a web-based interface to manage AFS instances.

Tokens
248.2K
Snippets
873
Records
1.1K
Agent score
62%

What's inside aigne-framework

  1. Overview of AIGNE Observability features

    main

    AIGNE Observability is an OpenTelemetry-based visualization tool designed to monitor AI Agent runtime status. Key capabilities include:

    • Real-time Monitoring: Visualize trace data and call chains.
    • Precise Positioning: Identify AIGNE internal workflows and locate issues.
    • Comprehensive Data: Collection of both Trace and Log data.
    • Flexible Deployment: Can be run as a local CLI service or integrated into a Blocklet environment.

  2. Overview of the Bash Agent

    main

    The Bash Agent enables secure execution of bash scripts within a sandboxed environment using Anthropic's Sandbox Runtime. It provides controlled access to system commands, network resources, and the filesystem while returning comprehensive execution results including stdout, stderr, and exitCode.

    Execution Modes

    • Sandboxed Mode (default): Scripts run with security restrictions (network, filesystem, and command permissions).
    • Direct Mode (sandbox: false): Scripts run without restrictions. This is required for Windows users as sandbox mode is not supported on Windows.

    ⚠️ Windows Platform Note: You must set sandbox: false in your configuration to use the Bash Agent on Windows.

  4. What is AIGNE Observability

    main
    AIGNE Observability is a visual monitoring tool designed to track Agent data flow. It is built on OpenTelemetry and supports the collection of both Trace and Log data. It can be deployed as a standalone service or used as an integrated module within the AIGNE runtime (which is integrated by default).

  5. Overview of AIGNE Framework

    main

    AIGNE Framework is a modular, multi-agent collaborative application development framework. It is designed to help developers organize, compose, and deploy AI capabilities through an agent-centric architecture.

    Key features include:

    • Multi-Agent Support: Designed for building collaborative workflows between multiple agents.
    • Modular Architecture: Functional modules can be composed on demand.
    • Type Safety: Built with TypeScript for development reliability.
    • Multi-Model Compatibility: Supports OpenAI, Claude, Gemini, Anthropic, and other mainstream AI services.
    • Protocol-Based Communication: Built-in support for the Model Context Protocol (MCP) to enable cross-system calls.
    • Blocklet Integration: Native support for one-click deployment on the Blocklet platform.

  7. Overview of AIGNE Observability

    main
    AIGNE Observability is a visual monitoring tool built on OpenTelemetry designed to track Agent data flow. It provides real-time visualization of trace data, call chains, and internal AIGNE workflows. It supports the collection of both Trace and Log data and can be used as a standalone service or integrated into the AIGNE runtime (which includes this module by default).

  8. AIGNE Default Template Overview

    main

    The AIGNE Default Template is a starter project for the AIGNE framework. It provides a pre-configured environment for building chat agents with built-in JavaScript code execution capabilities.

    The template consists of the following core files:

    • aigne.yaml: The main project configuration file. It defines the chat model and references the agents used in the project.
    • chat.yaml: Configuration for the chat agent, specifying its instructions and the skills it can utilize.
    • sandbox.js: A tool implementation for executing JavaScript code within agent conversations.
    • sandbox.test.js: A test suite used to verify the functionality of the JavaScript execution tool.

  9. AIGNE CLI core workflows

    main

    The AIGNE CLI supports four primary workflows for managing AI Agents:

    1. Project Creation: Initialize new AIGNE projects (see create command).
    2. Agent Execution: Run and configure agents (see run command).
    3. Testing: Execute the built-in testing framework to ensure agent quality and stability (see test command).
    4. Deployment: Deploy agents as Model Context Protocol (MCP) servers or API services (see serve-mcp command).

  10. Understand the AIGNE Default Template structure

    main

    The AIGNE default template provides a foundation for building chat agents with JavaScript code execution capabilities. The project is organized around the following core files:

    • aigne.yaml: The central project configuration file. It defines the chat model and references the agents used in the project.
    • chat.yaml: Configuration for the chat agent, specifying its instructions and the skills it can utilize.
    • sandbox.js: A tool implementation that allows for executing JavaScript code within conversations.
    • sandbox.test.js: A test suite used to verify that the code execution tool (sandbox) is functioning correctly.

  11. Supported Model Providers in AIGNE

    main

    The AIGNE framework supports multiple chat model providers, allowing you to switch between different AI capabilities. Supported providers include:

    • OpenAI: GPT series models via @aigne/openai
    • Anthropic: Claude series models via @aigne/anthropic
    • Google Gemini: Gemini series models via @aigne/gemini
    • AWS Bedrock: Various foundation models (e.g., Amazon nova) via @aigne/bedrock
    • Ollama: Local self-hosted open source models via @aigne/ollama
    • OpenRouter: Unified API access to multiple models via @aigne/open-router
    • DeepSeek: DeepSeek models via @aigne/deepseek
    • XAI: X.AI's Grok models via @aigne/xai

  12. Core capabilities of @aigne/cli

    main

    The @aigne/cli provides several key features to accelerate agent development workflows:

    • Project Scaffolding: Use aigne create to quickly generate a new AIGNE project with a predefined file structure and configuration.
    • Local Agent Execution: Use aigne run to execute agents in a local chat loop, allowing for rapid testing and debugging through interaction.
    • Automated Testing: Use aigne test to run unit and integration tests to ensure agent robustness and reliability.
    • MCP Server Integration: Start agents as Model Context Protocol (MCP) servers to enable integration with external systems.
    • Observability: Use aigne observe to start a local server for viewing and analyzing agent execution traces and performance data.
    • Multi-model Support: Seamlessly switch between various AI model providers such as OpenAI, Claude, and XAI.