llmagent

LLMAgent Example

This example demonstrates how to use the LLMAgent implementation with an interactive chat interface.

What is LLMAgent?

The LLMAgent is a concrete implementation of the agent.Agent interface that uses language models (LLMs) to generate responses. It leverages the llmflow package internally to handle the execution flow.

Key Features

• 🔄 Interactive Chat: Multi-turn conversation interface with streaming responses
• 🌊 Streaming Control: Choose between real-time streaming or batch responses
• 🚀 Simple Interface: Clean, focused chat experience
• 🔧 Implements agent.Agent interface: Provides a Run method that accepts a context and invocation, returning a channel of events
• ⚙️ Configurable: Supports custom channel buffer sizes, request processors, and response processors
• 🌊 Flow-based execution: Uses the llmflow package for handling LLM interactions
• ⚡ Event-driven: Communicates through events that can include LLM responses, errors, and metadata

Prerequisites

• Go 1.21 or later
• Valid OpenAI API key (or compatible API endpoint)

Environment Variables

Variable	Description	Default Value
OPENAI_API_KEY	API key for the model service (required, automatically read by OpenAI SDK)	``
OPENAI_BASE_URL	Base URL for the model API endpoint (automatically read by OpenAI SDK)	https://api.openai.com/v1

Note: OPENAI_API_KEY and OPENAI_BASE_URL are automatically read by the OpenAI SDK. You don't need to manually read these environment variables in your code. The SDK handles this automatically when creating the client.

Command Line Arguments

Argument	Description	Default Value
-model	Name of the model to use	deepseek-v4-flash
-streaming	Enable streaming mode	true

Usage

Basic Chat

cd examples/llmagent
export OPENAI_API_KEY="your-api-key-here"
go run . # or: go run main.go

Custom Model

export OPENAI_API_KEY="your-api-key"
go run . -model gpt-4o # or: go run main.go -model gpt-4o

Response Modes

Choose between streaming and non-streaming responses:

# Default streaming mode (real-time character output)
go run . # or: go run main.go

# Non-streaming mode (complete response at once)
go run . -streaming=false # or: go run main.go -streaming=false

Chat Interface

The interface is simple and intuitive:

🚀 Interactive Chat with LLMAgent
Model: deepseek-v4-flash
Streaming: true
==================================================
✅ Chat ready!

💡 Commands:
   /exit     - End the conversation

👤 You: Hello! Can you help me with a question?
🤖 Assistant: Of course! I'd be happy to help. What's your question?

👤 You: What is the capital of France?
🤖 Assistant: The capital of France is **Paris**! 🇫🇷

It's a beautiful city known for its iconic landmarks like the Eiffel Tower, the Louvre Museum, and Notre-Dame Cathedral. Paris is also famous for its art, fashion, and delicious cuisine. Have you ever been, or are you planning a visit?

👤 You: /exit
👋 Goodbye!

Architecture

The LLMAgent follows this architecture:

LLMAgent
├── Implements agent.Agent interface
├── Uses llmflow.Flow internally
├── Configurable via Options struct
└── Returns events via channel

Components:
- **🤖 Model**: OpenAI-compatible model for LLM calls
- **⚙️ RequestProcessors**: Process requests before sending to LLM
- **⚡ ResponseProcessors**: Process responses after receiving from LLM
- **🌊 Flow**: Handles the execution logic and event generation

Advanced Usage

To create a more functional LLMAgent, you would typically:

1. 🔧 Add Request Processors: These prepare the LLM request with appropriate messages
2. ⚡ Add Response Processors: These handle the LLM responses and can trigger additional actions
3. ⚙️ Configure Buffer Sizes: Optimize for your specific throughput requirements

Example with function options:

agent := llmagent.New(
    "advanced-agent",
    llmagent.WithModel(modelInstance),
    llmagent.WithDescription("An advanced AI assistant"),
    llmagent.WithInstruction("Your agent instruction"),
    llmagent.WithSystemPrompt("Your system prompt"),
    llmagent.WithTools(tools),
    llmagent.WithPlanner(planner),
)