Skip to content

Release Notes - GoLLM v0.6.0

Release Date: 2025-12-14 Base Version: v0.5.1 (commit d8763ee)

🎉 Overview

Version 0.6.0 is a major feature release that adds X.AI Grok provider support, implements Anthropic streaming, introduces comprehensive testing infrastructure, and enhances the canonical response structure to preserve provider-specific metadata.

Summary: - ✨ New X.AI Grok provider with full streaming support - ✨ Complete Anthropic streaming implementation - 🧪 Comprehensive test suite with 44 tests (unit + integration) - 📊 Provider metadata preservation in responses - 🏗️ Mock infrastructure for testing - 📚 Enhanced documentation and examples - 🔧 Code quality improvements

Statistics: - Files Changed: 22 files - Lines Added: 3,041 insertions - Lines Removed: 61 deletions - New Files: 12 files - Test Coverage: 44 automated tests

✨ New Features

1. X.AI Grok Provider Support

Complete implementation of X.AI's Grok API with OpenAI-compatible interface.

New Files: - providers/xai/xai.go - HTTP client with SSE streaming (204 lines) - providers/xai/types.go - Request/response types (68 lines) - providers/xai/adapter.go - Provider interface implementation (159 lines) - providers/xai/integration_test.go - Integration tests (181 lines) - examples/xai/main.go - Usage examples (161 lines)

Supported Models:

Grok 4.1 (Latest - November 2025): - grok-4-1-fast-reasoning - Best tool-calling model with 2M context window - grok-4-1-fast-non-reasoning - Instant responses with 2M context window

Grok 4 (July 2025): - grok-4-0709 - Flagship model with 256K context - grok-4-fast-reasoning - Fast reasoning with 2M context - grok-4-fast-non-reasoning - Fast non-reasoning with 2M context - grok-code-fast-1 - Coding-optimized model with 256K context

Grok 3: - grok-3 - Grok 3 model - grok-3-mini - Smaller, faster variant

Grok 2: - grok-2-1212 - Grok 2 (December 2024) - grok-2-vision-1212 - Grok 2 with vision capabilities

Example Usage: 

client, err := gollm.NewClient(gollm.ClientConfig{
    Provider: gollm.ProviderNameXAI,
    APIKey:   os.Getenv("XAI_API_KEY"),
})

Constants Added: - ProviderNameXAI - Provider identifier - EnvVarXAIAPIKey - Environment variable constant - Grok 4.1 Models: ModelGrok4_1FastReasoning, ModelGrok4_1FastNonReasoning - Grok 4 Models: ModelGrok4_0709, ModelGrok4FastReasoning, ModelGrok4FastNonReasoning, ModelGrokCodeFast1 - Grok 3 Models: ModelGrok3, ModelGrok3Mini - Grok 2 Models: ModelGrok2_1212, ModelGrok2_Vision

2. Anthropic Streaming Implementation

Full native streaming support for Anthropic Claude using Server-Sent Events (SSE).

Files Modified: - providers/anthropic/anthropic.go - Added SSE streaming client (+126 lines) - providers/anthropic/adapter.go - Streaming adapter implementation (+174 lines) - providers/anthropic/types.go - Streaming event types (+32 lines)

New Streaming Types: - StreamEvent - SSE event container - StreamDelta - Delta content for streaming - StreamMessage - Message metadata - StreamUsage - Streaming usage information

Example: 

stream, err := client.CreateChatCompletionStream(ctx, &gollm.ChatCompletionRequest{
    Model: gollm.ModelClaude3Sonnet,
    Messages: []gollm.Message{
        {Role: gollm.RoleUser, Content: "Write a haiku"},
    },
})
defer stream.Close()

for {
    chunk, err := stream.Recv()
    if err == io.EOF {
        break
    }
    if len(chunk.Choices) > 0 && chunk.Choices[0].Delta != nil {
        fmt.Print(chunk.Choices[0].Delta.Content)
    }
}

Example Added: - examples/anthropic_streaming/main.go - Complete streaming demonstrations (158 lines)

3. Provider Metadata Preservation

New ProviderMetadata field in response structures to preserve provider-specific information.

Files Modified: - provider/types.go - Added ProviderMetadata map[string]any to: - ChatCompletionResponse (line 74) - ChatCompletionChunk (line 102)

Anthropic Metadata Preserved: - anthropic_type - Response type - anthropic_role - Response role - anthropic_content - Full content array (preserves multi-block responses) - anthropic_stop_reason - Stop reason - anthropic_event_type - Streaming event types - anthropic_delta - Delta objects - anthropic_usage - Usage information

Example Usage: 

response, err := client.CreateChatCompletion(ctx, req)

// Access standard unified fields
content := response.Choices[0].Message.Content

// Access Anthropic-specific metadata
if metadata := response.ProviderMetadata; metadata != nil {
    if fullContent, ok := metadata["anthropic_content"].([]anthropic.Content); ok {
        // Access all content blocks, not just first
        for _, block := range fullContent {
            fmt.Printf("Block [%s]: %s\n", block.Type, block.Text)
        }
    }
}

Benefits: - ✅ Backward compatible - existing code continues working - ✅ Provider-agnostic - any provider can add metadata - ✅ No data loss - preserves full provider responses - ✅ Optional - field omitted when empty

4. Comprehensive Testing Infrastructure

Complete test suite with unit tests, integration tests, and mock implementations.

New Test Files: - client_test.go - Client and memory integration tests (441 lines) - memory_test.go - Conversation memory tests (314 lines) - providers/anthropic/adapter_test.go - Anthropic unit tests (270 lines) - providers/anthropic/integration_test.go - Anthropic API tests (302 lines) - providers/openai/integration_test.go - OpenAI API tests (181 lines) - providers/xai/integration_test.go - X.AI API tests (181 lines) - testing/mock_kvs.go - Mock KVS implementation (102 lines)

Test Statistics: - Total Tests: 44 automated tests - Unit Tests: ~30 tests (run without API keys) - Integration Tests: ~14 tests (conditional on API keys)

Test Categories:

  1. Client Tests (client_test.go)
  2. Client creation and configuration
  3. Chat completion (sync and streaming)
  4. Memory-aware completions
  5. Conversation management

  6. Provider injection

  7. Memory Tests (memory_test.go)

  8. Load/save conversations
  9. Append messages
  10. Max message limits
  11. System message preservation
  12. Metadata handling

  13. Conversation deletion

  14. Provider Tests (Anthropic, OpenAI, X.AI)

  15. Message conversion
  16. Request/response mapping
  17. Streaming adapters
  18. Error handling
  19. Real API integration (conditional)

Integration Test Pattern: 

func TestAnthropicIntegration_Streaming(t *testing.T) {
    apiKey := os.Getenv("ANTHROPIC_API_KEY")
    if apiKey == "" {
        t.Skip("Skipping integration test: ANTHROPIC_API_KEY not set")
    }
    // Test actual API calls...
}

Running Tests: 

# Run all unit tests (no API keys required)
go test ./... -short

# Run with coverage
go test ./... -short -cover

# Run integration tests (requires API keys)
ANTHROPIC_API_KEY=key go test ./providers/anthropic -v
OPENAI_API_KEY=key go test ./providers/openai -v
XAI_API_KEY=key go test ./providers/xai -v

Mock Infrastructure: - testing/mock_kvs.go - In-memory KVS for testing conversation memory without external dependencies - Mock providers for unit testing client logic

🔧 Improvements

Code Quality

  1. Type Modernization
  2. Replaced interface{} with any throughout codebase
  3. Updated in provider/types.go, client.go, providers/anthropic/adapter.go

  4. Improves readability and follows Go 1.18+ conventions

  5. Lint Fixes

  6. Fixed all golangci-lint issues
  7. Removed ineffectual assignments
  8. Removed unused struct fields
  9. Fixed gofmt issues

  10. Result: 0 lint issues

  11. Memory Function Improvements

  12. Updated memory.go with better type safety (18 lines modified)

  13. X.AI Grok 4 Model Support ⭐ NEW

  14. Added 6 new Grok 4/4.1 model constants
  15. Grok 4.1 Fast: Latest with 2M context window and tool calling
  16. Grok 4 Fast: 2M context with reasoning/non-reasoning modes
  17. Grok Code Fast: Optimized for coding tasks (256K context)
  18. Updated examples and tests to use grok-4-1-fast-reasoning

Documentation

  1. README.md Updates (166 lines modified)
  2. Added X.AI provider documentation
  3. Updated architecture diagram with xai/ provider
  4. Added comprehensive Testing section
  5. Updated model listings for all providers:
    • OpenAI: Added GPT-5, GPT-4.1 models
    • Anthropic: Added Claude-Opus-4.1, Claude-Opus-4, Claude-Sonnet-4, Claude-3.7-Sonnet, Claude-3.5-Haiku
    • Gemini: Updated to current models
  6. Enhanced Contributing section with test requirements

  7. Added integration test examples with X.AI

  8. Provider Support Table Updated 

    | Provider   | Models                           | Features                          |
|------------|----------------------------------|-----------------------------------|
| OpenAI     | GPT-5, GPT-4.1, GPT-4o, ...     | Chat, Streaming, Functions        |
| Anthropic  | Claude-Opus-4.1, Sonnet-4, ...  | Chat, Streaming, System messages  |
| Gemini     | Gemini-2.5-Pro, 2.5-Flash, ...  | Chat, Streaming                   |
| Bedrock    | Claude models, Titan models      | Chat, Multiple model families     |
| X.AI       | Grok-4.1-Fast, Grok-4, Grok-4-Fast, Grok-Code-Fast, Grok-3, Grok-2 | Chat, Streaming, 2M context, Tool calling |
| Ollama     | Llama 3, Mistral, Gemma, ...    | Chat, Streaming, Local inference  |

  9. New Examples

  10. examples/anthropic_streaming/main.go - 3 Anthropic streaming scenarios
  11. examples/xai/main.go - 3 X.AI usage examples

📊 Statistics

Code Additions by Component

Component Lines Added Description
X.AI Provider 612 Complete provider implementation
Anthropic Streaming 332 Native SSE streaming support
Test Infrastructure 1,689 Unit + integration + mock tests
Examples 319 Usage demonstrations
Documentation 166 README and inline docs
Total 3,041 Net additions to codebase

File Distribution

  • Provider Code: 7 files (612 lines for X.AI, 332 for Anthropic streaming)
  • Tests: 6 files (1,689 lines)
  • Examples: 2 files (319 lines)
  • Core Changes: 5 files (provider/types.go, client.go, constants.go, providers.go, memory.go)
  • Mock Infrastructure: 1 file (102 lines)

🔄 Breaking Changes

None - This release is fully backward compatible.

All changes are additive: - New provider (X.AI) doesn't affect existing providers - ProviderMetadata field is optional (omitempty) - Existing APIs unchanged - Test additions don't affect runtime behavior

🐛 Bug Fixes

  1. Anthropic Streaming
  2. Previously unimplemented streaming now fully functional
  3. Proper SSE parsing with event type handling

  4. Correct message assembly from multiple events

  5. Lint Issues

  6. Fixed ineffectual assignment in anthropic/anthropic.go:189

  7. Removed unused field streamChunkIndex in client_test.go:20

  8. Model Updates

  9. Updated X.AI examples and tests to use latest grok-4-1-fast-reasoning model
  10. Added support for Grok 4 and Grok 4.1 model families (6 new model constants)
  11. Added deprecation markers for grok-beta and grok-vision-beta

🚀 Migration Guide

Upgrading from v0.5.1

No code changes required - simply update your dependency:

go get -u github.com/grokify/gollm@v0.6.0

Using New Features

1. Add X.AI Grok Support: 

client, err := gollm.NewClient(gollm.ClientConfig{
    Provider: gollm.ProviderNameXAI,
    APIKey:   os.Getenv("XAI_API_KEY"),
})

2. Use Anthropic Streaming: 

stream, err := client.CreateChatCompletionStream(ctx, &gollm.ChatCompletionRequest{
    Model: gollm.ModelClaude3Sonnet,
    Messages: []gollm.Message{
        {Role: gollm.RoleUser, Content: "Hello!"},
    },
})

3. Access Provider Metadata: 

response, err := client.CreateChatCompletion(ctx, req)
if metadata := response.ProviderMetadata; metadata != nil {
    // Access provider-specific data
    fmt.Printf("Metadata: %+v\n", metadata)
}

4. Run Tests: 

# Unit tests only
go test ./... -short

# Integration tests (requires API keys)
export ANTHROPIC_API_KEY=your-key
export OPENAI_API_KEY=your-key
export XAI_API_KEY=your-key
go test ./... -v

📝 Commits Since v0.5.1

b9ed829 style: `any`: update from `interface{}`
a509740 feat: `provider`: add `ChatCompletionChunk.ProviderMetadata`
803e5bb docs: `README.md`: update
0c438f0 chore(lint): `golangci-lint`: fix various
65917c8 feat: `providers/xai`: add support for xAI
f21d267 chore(lint): `golangci-lint`: fix `gofmt`
6fd5654 chore(lint): `golangci-lint`: fix `gofmt`
536d73c docs: `README.md`: update for Anthropic streaming
6514343 tests: add
5f430f5 feat: `providers/anthropic`: add streaming support

🙏 Acknowledgments

This release represents a significant expansion of GoLLM's capabilities with: - Complete X.AI Grok provider integration - Production-ready Anthropic streaming - Enterprise-grade test coverage - Enhanced metadata preservation

📚 Resources

  • Documentation: README
  • Examples: See /examples directory
  • Tests: Run go test ./... -v to see all test scenarios
  • Issues: GitHub Issues

Full Changelog: v0.5.1...v0.6.0