Skip to content

README.md

Latest commit

 

History

History
238 lines (170 loc) · 7.36 KB

README.md

File metadata and controls

238 lines (170 loc) · 7.36 KB

ReasonKit Integrations

Connect ReasonKit to any LLM provider or CLI tool "Turn Prompts into Protocols - Anywhere"

ReasonKit AI Integration Options: Claude, Gemini, OpenAI, Cursor, VS Code, Any LLM

Universal Compatibility: ReasonKit integrates seamlessly with Claude, Gemini, OpenAI, Cursor, VS Code, and any LLM provider. The same structured reasoning protocols work across all platforms, giving you flexibility without vendor lock-in.

LLM Provider Integration Guides

Comprehensive guides for integrating ReasonKit with major LLM providers:

Provider Guide Best For Quick Start
Anthropic (Claude) ANTHROPIC_INTEGRATION.md Complex reasoning, extended thinking, safety-critical rk think --provider anthropic "query"
OpenAI (GPT) OPENAI_INTEGRATION.md General reasoning, JSON mode, embeddings rk think --provider openai "query"
Google (Gemini) GOOGLE_INTEGRATION.md Long context (2M), multimodal, documents rk think --provider gemini "query"
Groq GROQ_INTEGRATION.md Ultra-fast inference, rapid iteration rk think --provider groq "query"
OpenRouter OPENROUTER_INTEGRATION.md 300+ models, fallback routing, cost optimization rk think --provider openrouter "query"

Quick Comparison

Speed vs Quality vs Cost

Provider Speed Quality Cost Best Profile
Groq Fastest Good Lowest --quick
Gemini Flash Very Fast Good Very Low --quick, --balanced
OpenAI GPT-4o Fast Excellent Medium --balanced
Claude Sonnet Fast Excellent Medium --balanced, --deep
Claude Opus Medium Best High --deep, --paranoid
OpenAI o1 Slow Excellent High --deep, --paranoid

Feature Matrix

Feature Anthropic OpenAI Gemini Groq OpenRouter
Max Context 200K 200K 2M 128K Varies
Extended Thinking Yes Yes (o1) Yes No Via models
JSON Mode No Yes Yes Yes Via models
Vision Yes Yes Yes No Via models
Audio No Yes Yes Yes Via models
Video No No Yes No Via models
Embeddings No Yes Yes No Via models
Free Tier No No Yes Limited Yes

Environment Setup (All Providers)

Set all API keys for maximum flexibility:

# Core providers
export ANTHROPIC_API_KEY="sk-ant-..."    # Anthropic Claude
export OPENAI_API_KEY="sk-..."           # OpenAI GPT
export GEMINI_API_KEY="..."              # Google Gemini
export GROQ_API_KEY="gsk_..."            # Groq
export OPENROUTER_API_KEY="sk-or-..."    # OpenRouter (300+ models)

# Additional providers
export XAI_API_KEY="xai-..."             # xAI Grok
export MISTRAL_API_KEY="..."             # Mistral AI
export DEEPSEEK_API_KEY="..."            # DeepSeek

Default Configuration

~/.ReasonKit/config.toml

# Default provider for rk think
[thinktool]
default_provider = "anthropic"
default_model = "claude-sonnet-4"

# Provider-specific defaults
[providers.anthropic]
default_model = "claude-sonnet-4"

[providers.openai]
default_model = "gpt-4o"

[providers.gemini]
default_model = "gemini-2.0-flash"

[providers.groq]
default_model = "llama-3.3-70b-versatile"

[providers.openrouter]
default_model = "anthropic/claude-sonnet-4"

Usage Patterns

Single Provider

# Use specific provider
rk think --provider anthropic "Analyze this code"
rk think --provider openai "Evaluate this design"
rk think --provider gemini "Summarize this document"
rk think --provider groq "Quick review"

Provider + Model

# Specify exact model
rk think --provider anthropic --model claude-opus-4 "Deep analysis"
rk think --provider openai --model o1 "Complex reasoning"
rk think --provider gemini --model gemini-1.5-pro "Long document"
rk think --provider groq --model llama-3.1-405b-reasoning "Hard problem"

Profile-Based Selection

# Profiles auto-select appropriate models
rk think --profile quick "Fast check"     # Uses fastest available
rk think --profile balanced "Standard"    # Uses balanced model
rk think --profile deep "Thorough"        # Uses reasoning model
rk think --profile paranoid "Critical"    # Uses best model

CLI Tool Integration

ReasonKit can act as the default reasoning layer inside popular CLI agents:

Tool Injection Method Bypass
claude --append-system-prompt --no-rk
gemini Prompt prefix --no-rk
codex Prompt prefix --no-rk
opencode Prompt prefix --no-rk
cursor-agent Prompt prefix --no-rk
copilot via gh copilot --no-rk

Install CLI Defaults

bash reasonkit-core/scripts/install_cli_defaults.sh
source ~/.zshrc

Central Config

All defaults are controlled by reasonkit-core/config/cli_defaults.toml:

export RK_CONFIG="$HOME/RK-PROJECT/reasonkit-core/config/cli_defaults.toml"

Example:

[defaults]
profile = "balanced"
protocol_dir = "protocols/cli"

[tools.codex]
profile = "paranoid"

Override Per-Session

export RK_PROFILE=quick
# or
export RK_PROFILE=paranoid

Bypass ReasonKit

claude --no-rk  # Run without ReasonKit injection

Protocol Files

Protocols are stored in reasonkit-core/protocols/cli/:

  • balanced.md
  • quick.md
  • paranoid.md

Cost Management

Budget Controls

# Set per-query budget
rk think "Expensive query" --budget "$1.00"

# Set time budget
rk think "Time-limited" --budget "60s"

# Set token budget
rk think "Token-limited" --budget "5000t"

Cost Tracking

# View usage
rk metrics cost --period day
rk metrics cost --period month
rk metrics cost --provider anthropic

Additional Resources

ReasonKit Integrations | v1.0.0 | Apache 2.0 "See How Your AI Thinks"