Anyi AI Agent

Anyi - Open Source Autonomous AI Agent Framework

| English | 中文 |

Anyi is a powerful autonomous AI agent framework that helps you build AI solutions that seamlessly integrate with real-world workflows through unified LLM interfaces, robust validation mechanisms, and flexible workflow systems.

Key Features

• Universal LLM Access - Connect to multiple LLM providers (OpenAI, Anthropic, etc.) through a consistent API
• Powerful Workflow System - Chain steps together with validation and automatic retries for robust AI pipelines
• Configuration-Driven Development - Define workflows and clients through code or external config files (YAML, JSON, TOML)
• Multimodal Support - Send text and images to compatible models simultaneously
• Go Template Integration - Use Go's template engine to generate dynamic prompts

When to Use Anyi

Anyi is ideal for:

• AI Application Development - Build production-ready AI services with reliable error handling and retries
• Multi-LLM Applications - Create solutions that leverage different models for different tasks based on cost or capability
• DevOps Integration - Connect AI capabilities with your existing systems through command executors and API integrations
• Rapid Prototyping - Configure complex AI workflows through config files without changing code
• Enterprise Solutions - Maintain separation of code and configuration for production deployment across environments

Supported LLM Providers

• OpenAI - GPT series models
• Azure OpenAI - Microsoft-hosted OpenAI models
• Anthropic - Claude series models
• Zhipu AI - GLM series models
• Alibaba Cloud Lingji - Qwen series models
• Ollama - Local deployment of open-source models (Llama, Qwen, etc.)
• DeepSeek - DeepSeek models
• SiliconCloud - SiliconFlow models

Quick Start

Installation

go get -u github.com/jieliu2000/anyi

Requires Go 1.20 or higher

Basic Usage

package main

import (
	"log"
	"os"

	"github.com/jieliu2000/anyi"
	"github.com/jieliu2000/anyi/llm/openai"  // Import your preferred provider
	"github.com/jieliu2000/anyi/llm/chat"
)

func main() {
	// Create client - just change imports and config to use different providers
	config := openai.DefaultConfig("gpt-4")
	config.APIKey = os.Getenv("OPENAI_API_KEY")

	client, err := anyi.NewClient("gpt4", config)
	if err != nil {
		log.Fatalf("Failed to create client: %v", err)
	}

	// Send chat request
	messages := []chat.Message{
		{Role: "user", Content: "How many countries are in Africa?"},
	}

	response, _, err := client.Chat(messages, nil)
	if err != nil {
		log.Fatalf("Chat failed: %v", err)
	}

	log.Printf("Response: %s", response.Content)
}

Creating Workflows

Using Code

package main

import (
	"log"
	"os"
	"github.com/jieliu2000/anyi"
	"github.com/jieliu2000/anyi/llm/openai"
)

func main() {
	// Create client
	config := openai.DefaultConfig("gpt-4")
	config.APIKey = os.Getenv("OPENAI_API_KEY")
	client, err := anyi.NewClient("gpt4", config)
	if err != nil {
		log.Fatalf("Failed to create client: %v", err)
	}

	// Create a two-step workflow
	step1, _ := anyi.NewLLMStepWithTemplate(
		"Generate a short story about {{.Text}}",
		"You are a creative fiction writer.",
		client,
	)
	step1.Name = "story_generation"

	step2, _ := anyi.NewLLMStepWithTemplate(
		"Create an engaging title for the following story:\n\n{{.Text}}",
		"You are an editor skilled at creating titles.",
		client,
	)
	step2.Name = "title_creation"

	// Create and register the flow
	myFlow, _ := anyi.NewFlow("story_flow", client, *step1, *step2)
	anyi.RegisterFlow("story_flow", myFlow)

	// Run the workflow
	result, _ := myFlow.RunWithInput("a detective in future London")

	log.Printf("Title: %s", result.Text)
}

Using Configuration Files

Anyi supports configuration-driven development, allowing you to define LLM clients and workflows in external files:

# config.yaml
clients:
  - name: "gpt4"
    type: "openai"
    config:
      model: "gpt-4"
      apiKey: "$OPENAI_API_KEY" # References environment variable

  - name: "ollama"
    type: "ollama"
    config:
      model: "llama3"
      baseURL: "http://localhost:11434" # Default Ollama server address

flows:
  - name: "story_flow"
    clientName: "gpt4" # Default client for the flow
    steps:
      - name: "story_generation"
        executor:
          type: "llm"
          withconfig:
            template: "Generate a short story about {{.Text}}"
            systemMessage: "You are a creative fiction writer."
        maxRetryTimes: 2

      - name: "title_creation"
        executor:
          type: "llm"
          withconfig:
            template: "Create an engaging title for the following story:\n\n{{.Text}}"
            systemMessage: "You are an editor skilled at creating titles."
        clientName: "ollama" # Override client for this step

Load and use this configuration:

package main

import (
	"log"
	"github.com/jieliu2000/anyi"
)

func main() {
	// Load configuration from file
	err := anyi.ConfigFromFile("./config.yaml")
	if err != nil {
		log.Fatalf("Failed to load config: %v", err)
	}

	// Get and run the configured flow
	flow, err := anyi.GetFlow("story_flow")
	if err != nil {
		log.Fatalf("Failed to get flow: %v", err)
	}

	result, err := flow.RunWithInput("a detective in future London")
	if err != nil {
		log.Fatalf("Flow execution failed: %v", err)
	}

	log.Printf("Result: %s", result.Text)
}

Built-in Components

Executors

• LLMExecutor - Sends prompts to large language models
• SetContextExecutor - Modifies workflow context
• ConditionalFlowExecutor - Branches based on conditions
• RunCommandExecutor - Executes system commands
• MCPExecutor - Interfaces with Model Control Protocol to access external models, resources, and tools

Validators

• StringValidator - Checks text via regex or equality
• JsonValidator - Ensures output is valid JSON

Documentation

Getting Started

• Installation and Setup - System requirements and installation
• Quick Start Guide - Your first Anyi application
• Basic Concepts - Understanding clients, flows, and executors

Tutorials

• Working with LLM Clients - Comprehensive guide to all supported providers
• Building Workflows - Creating complex AI workflows
• Configuration Management - Using config files and environment variables
• Multimodal Applications - Working with text and images

How-To Guides

• Provider Setup - Detailed setup for each LLM provider
• Error Handling - Best practices for robust applications
• Performance Optimization - Speed and cost optimization
• Web Integration - Using Anyi with web frameworks

Reference

• API Reference - Complete API documentation
• Configuration Schema - All configuration options
• Built-in Components - Executors and validators reference

Advanced Topics

• Custom Executors - Building your own executors
• Security Best Practices - Securing your AI applications
• Production Deployment - Production considerations

Contributing

Contributions welcome! Anyi is under active development, and your feedback helps make it better for everyone.

License

Anyi is licensed under the Apache License 2.0.