Skip to content

Release Notes - OmniLLM v0.10.0

Release Date: 2026-01-04 Base Version: v0.9.0

Overview

Version 0.10.0 adds configurable HTTP client timeout support and Claude 4.5 model constants. The timeout configuration is essential for reasoning models that may take longer to respond.

Summary:

  • Configurable Timeout: New ClientConfig.Timeout field for setting HTTP client timeout
  • Claude 4.5 Models: Added model constants for Claude Opus 4.5, Sonnet 4.5, and Haiku 4.5
  • HTTP Client Refactor: Unified HTTP client creation with getHTTPClient() helper

New Features

1. Configurable HTTP Client Timeout

Added Timeout field to ClientConfig for configuring HTTP client timeout. This is particularly important for reasoning models (like xAI Grok reasoning models) that may need more time to complete.

Usage: 

import (
    "time"
    "github.com/agentplexus/omnillm"
)

client, err := omnillm.NewClient(omnillm.ClientConfig{
    Provider: omnillm.ProviderNameXAI,
    APIKey:   os.Getenv("XAI_API_KEY"),
    Timeout:  300 * time.Second, // 5 minutes for reasoning models
})

Behavior:

  • If Timeout is set and HTTPClient is nil, a new HTTP client is created with the specified timeout
  • If HTTPClient is provided, it takes precedence (for custom transports with retry logic, etc.)
  • If neither is set, providers use their default timeouts

Recommendation: Use 300 * time.Second (5 minutes) for reasoning models.

2. Claude 4.5 Model Constants

Added model constants for the Claude 4.5 family:

import "github.com/agentplexus/omnillm/models"

// Available Claude 4.5 models
models.ClaudeOpus4_5    // "claude-opus-4-5-20251101"
models.ClaudeSonnet4_5  // "claude-sonnet-4-5-20250929"
models.ClaudeHaiku4_5   // "claude-haiku-4-5-20251001"

3. Documentation

Added Marp presentation and GitHub Pages HTML documentation for the project.

Improvements

HTTP Client Handling Refactor

The HTTP client creation logic has been refactored with a new getHTTPClient() helper function that:

  1. Returns the custom HTTPClient if provided
  2. Creates a new client with Timeout if specified
  3. Returns nil to let providers use their defaults

This provides a unified and consistent HTTP client handling across all providers (OpenAI, Anthropic, xAI, Ollama).

Upgrade Guide

From v0.9.0

No breaking changes. To use the new timeout feature:

// Before (provider default timeout)
client, err := omnillm.NewClient(omnillm.ClientConfig{
    Provider: omnillm.ProviderNameXAI,
    APIKey:   apiKey,
})

// After (custom timeout for reasoning models)
client, err := omnillm.NewClient(omnillm.ClientConfig{
    Provider: omnillm.ProviderNameXAI,
    APIKey:   apiKey,
    Timeout:  300 * time.Second,
})

go get github.com/agentplexus/omnillm@v0.10.0
go mod tidy

Provider Default Timeouts

For reference, the default timeouts when no Timeout is configured:

Provider Default Timeout
OpenAI 30 seconds
Anthropic 30 seconds
xAI 60 seconds
Ollama 60 seconds
Gemini (SDK default)

Note: These defaults may be too short for reasoning models. Set Timeout: 300 * time.Second for longer-running inference tasks.