Release Notes: v0.17.0¶

Release Date: 2026-07-05

Highlights¶

Reasoning Support: Unified ReasoningEffort and ThinkingConfig across all thick providers

New Features¶

Re-exported Types and Constants¶

The following types and constants are now available directly from the omnillm package:

// Types
omnillm.ThinkingConfig

// Reasoning effort constants
omnillm.ReasoningEffortNone
omnillm.ReasoningEffortLow
omnillm.ReasoningEffortMedium
omnillm.ReasoningEffortHigh

// Thinking type constants (Anthropic)
omnillm.ThinkingTypeEnabled
omnillm.ThinkingTypeDisabled
omnillm.ThinkingTypeAdaptive

Reasoning Support¶

Control LLM reasoning depth with the unified ReasoningEffort or Thinking fields:

import "github.com/plexusone/omnillm"

effort := omnillm.ReasoningEffortHigh
response, err := client.CreateChatCompletion(ctx, &omnillm.ChatCompletionRequest{
    Model:           "o1-preview", // or claude-sonnet-4, gemini-2.0-flash-thinking-exp, etc.
    ReasoningEffort: &effort,
    Messages:        messages,
})

Provider Support¶

Provider	Module	Reasoning Support
OpenAI	omni-openai v0.5.0	`ReasoningEffort` for o1/o3 models
Anthropic	omni-anthropic v0.3.0	`ThinkingConfig` and `ReasoningEffort`
Gemini	omni-google v0.7.0	`ThinkingLevel` and `ReasoningEffort`
OpenRouter	omni-openrouter v0.2.0	Extended `ReasoningEffort` values
Bedrock	omni-aws v0.9.0	Partial (model-dependent)

ReasoningEffort Values¶

Constant	Value	Description
`ReasoningEffortNone`	`"none"`	Disable reasoning
`ReasoningEffortLow`	`"low"`	Light reasoning
`ReasoningEffortMedium`	`"medium"`	Balanced reasoning
`ReasoningEffortHigh`	`"high"`	Deep reasoning

Extended Thinking (Anthropic)¶

For Anthropic models, use ThinkingConfig for fine-grained control:

budget := int64(10000)
response, err := client.CreateChatCompletion(ctx, &omnillm.ChatCompletionRequest{
    Model: "claude-sonnet-4-6",
    Thinking: &omnillm.ThinkingConfig{
        Type:         omnillm.ThinkingTypeEnabled,
        BudgetTokens: &budget,
    },
    Messages: messages,
})

Dependencies¶

Module	Previous	New
omnillm-core	v0.17.0	v0.18.0
omni-openai	v0.4.1	v0.5.0
omni-anthropic	v0.2.2	v0.3.0
omni-google	v0.6.1	v0.7.0
omni-openrouter	v0.1.0	v0.2.0

Installation¶

go get github.com/plexusone/omnillm@v0.17.0

Documentation¶

Reasoning Feature Guide

Upgrade Notes¶

This is a backwards-compatible release. The new ReasoningEffort and Thinking fields are optional and do not affect existing code.