Skip to main content

Go Examples

Complete working example demonstrating TrainLoop LLM evaluation with Go.

Overview​

The Go examples demonstrate two core evaluation scenarios:

  • Code Generation: Testing LLM ability to write valid Python code
  • Letter Counting: Testing basic counting accuracy

Prerequisites​

  • Go 1.20+
  • OpenAI API key (or other supported LLM provider)

Quick Setup​

# Navigate to Go examples
cd examples/go

# Install Go dependencies
go mod tidy

# Note: If you encounter import issues with trainloop-llm-logging,
# the examples use the GitHub import path:
# import trainloop "github.com/trainloop/evals/sdk/go/trainloop-llm-logging"

# Create .env file with API keys
cp .env.example .env

Run Examples​

# Code generation example (evaluates if LLM can write valid code)
go run ai_request.go writes_valid_code.go

# Letter counting example (evaluates counting accuracy)  
go run ai_request.go counter_agent.go

# Run each script 3-4 times to collect samples
# Check collected data in trainloop/data/events/

Evaluate Results​

# Install TrainLoop CLI globally (recommended)
pipx install trainloop-cli

# Check that it installed correctly
trainloop --version

# Run evaluation
cd trainloop
trainloop eval

Key Components​

AI Request Utility (ai_request.go)​

package main

import (
    "context"
    "fmt"
    "os"
    
    "github.com/joho/godotenv"
    "github.com/sashabaranov/go-openai"
)

func init() {
    godotenv.Load()
}

func MakeAIRequest(prompt, model string, maxTokens int, extraHeaders map[string]string) (string, error) {
    if model == "" {
        model = "gpt-4"
    }

    client := openai.NewClient(os.Getenv("OPENAI_API_KEY"))
    
    req := openai.ChatCompletionRequest{
        Model:     model,
        MaxTokens: maxTokens,
        Messages: []openai.ChatCompletionMessage{
            {
                Role:    openai.ChatMessageRoleUser,
                Content: prompt,
            },
        },
    }

    resp, err := client.CreateChatCompletion(context.Background(), req)
    if err != nil {
        return "", fmt.Errorf("chat completion error: %v", err)
    }

    return resp.Choices[0].Message.Content, nil
}

TrainLoop Integration​

package main

import (
    "fmt"
    "log"
    
    trainloop "github.com/trainloop/evals/sdk/go/trainloop-llm-logging"
)

func init() {
    // CRITICAL: Call Collect() in init() BEFORE making AI requests
    trainloop.Collect()
}

func main() {
    // Tag requests for evaluation suites
    headers := trainloop.TrainloopTag("code-generation")
    response, err := MakeAIRequest(promptText, "gpt-4", 500, headers)
    if err != nil {
        log.Fatalf("Error: %v", err)
    }
    
    fmt.Printf("AI Response: %s\n", response)
}

Important Notes​

Import Path​

Use the GitHub import path for the TrainLoop SDK:

import trainloop "github.com/trainloop/evals/sdk/go/trainloop-llm-logging"

Initialization Order​

The trainloop.Collect() call must happen in init() before any AI requests:

func init() {
    trainloop.Collect()
}

Expected Output​

When you run the examples, you'll see:

AI Response: def factorial(n):
    if n < 0:
        raise ValueError("n must be a non-negative integer")
    elif n == 0 or n == 1:
        return 1
    else:
        return n * factorial(n-1)

Project Structure​

go/
├── ai_request.go          # AI request utility functions
├── writes_valid_code.go   # Code generation example
├── counter_agent.go       # Letter counting example
├── go.mod                 # Go module definition
├── go.sum                 # Go module checksums
└── trainloop/            # Evaluation configuration
    ├── eval/
    │   ├── metrics/      # Custom metrics
    │   └── suites/       # Evaluation suites
    └── data/
        ├── events/       # Collected LLM interactions
        └── results/      # Evaluation results

Next Steps​