08 SDK sdk guide

SDK Guide

Krnl-AI provides SDKs for both Python and .NET to build agent applications programmatically.

Python SDK

Installation

pip install krnlai

Quick Start

from krnlai import CognitiveAgent

agent = CognitiveAgent(safety_level="strict")
response = await agent.run("analyze this dataset")
print(response.output)

CognitiveAgent API

agent = CognitiveAgent(
    mode="auto",             # "auto" | "standalone" | "enterprise"
    safety_level="strict",   # "strict" | "relaxed"
    endpoint="",             # C# backend URL (enterprise mode)
    api_key="",              # C# backend API key
    max_iterations=10,
    enable_emotions=True,
    enable_learning=True,
)

Methods

Method	Description
run(command)	Execute a single command
run_command(envelope)	Execute a CommandEnvelope
stream(command)	Stream cognitive cycle events
close()	Clean up resources

Available Modules

Module	Description
krnlai.llm.openai	OpenAI LLM provider
krnlai.llm.anthropic	Anthropic LLM provider
krnlai.llm.ollama	Ollama LLM provider
krnlai.llm.google	Google Gemini provider
krnlai.llm.deepseek	DeepSeek provider
krnlai.llm.groq	Groq provider
krnlai.llm.openrouter	OpenRouter provider
krnlai.llm.mistral	Mistral AI provider
krnlai.llm.cohere	Cohere provider
krnlai.llm.together	Together AI provider
krnlai.core.safety	Safety checker and rules
krnlai.core.memory	Episodic, semantic, working, procedural, autobiographical memory
krnlai.core.emotion	VAD emotional model + pain/reward system
krnlai.core.policies	Policy engine
krnlai.core.risk	Risk scoring
krnlai.core.cognition	Cognitive cycle, metacognition, inner speech, latent planning
krnlai.core.consciousness	Operational consciousness, attention schema, global workspace, IIT
krnlai.core.world_model	Predictive world models (JEPA-based) for simulation
krnlai.core.causal	Graph-based causal reasoning and inference
krnlai.core.active_inference	Free-energy-based action selection
krnlai.core.dream	Dream simulation and consolidation
krnlai.core.continuous_learning	End-to-end continuous learning pipeline
krnlai.investigation	Causal investigation and root cause analysis
krnlai.integrations	LangChain, CrewAI, AutoGen, FastAPI
krnlai.enterprise	Enterprise client and stores

LLM Providers

from krnlai.llm.ollama import OllamaProvider

provider = OllamaProvider(model="llama3.1", endpoint="http://localhost:11434/v1")
response = await provider.chat("Hello!")

.NET SDK

Installation

dotnet add package KrnlAISdk

Quick Start

using KrnlAI;

var client = new KrnlAIClient();
var response = await client.RunAsync("analyze this dataset");
Console.WriteLine(response.Narration);

Models

Model	Description
AgentRunRequest	Request to run the agent
MemorySearchRequest	Request to search memory
EpisodeModels	Episode data models
HealthModels	Health check models
MetricsModels	Performance metrics models
GoalModels	Goal tracking models

Usage

// Search memory
var memoryResult = await client.SearchMemoryAsync("project decision");

// Check health
var health = await client.CheckHealthAsync();

// Get metrics
var metrics = await client.GetMetricsAsync();