Krims Code SDK

A lightweight, zero-dependency, universal developer SDK for integrating with the Krims Code ecosystem and querying a wide array of AI model providers.

Key Features

• Multi-Provider Support: Out-of-the-box wrappers for OpenAI, Google Gemini, Anthropic Claude, Cohere, Ollama (local), Groq, Together, Cerebras, Mistral, Fireworks, OpenRouter, DeepSeek, Perplexity, and xAI.
• Unified Chat Client: Perform stateless queries across different API providers with a single, simple .ask(prompt) method.
• Stateful Client-Side Sessions: Use KrimsSession to manage conversation histories locally. Chat histories are automatically mapped to the appropriate provider format.
• Zero-Config Integrations: Auto-detects standard environment variables (like OPENAI_API_KEY, GOOGLE_API_KEY, etc.).
• TypeScript Support: Native .d.ts declarations included for autocomplete and code validation.
• Zero Dependencies: Lightweight and fast, using native fetch and standard web APIs.

Installation

npm install krims-code-sdk

Quick Start

1. Zero-Config (Using local Krims Chatbot Server)

By default, the SDK targets the local Krims Chatbot backend at http://localhost:3000.

import { createKrimsClient } from 'krims-code-sdk';

const client = createKrimsClient();

// Call local health check
const health = await client.health();
console.log('Status:', health);

// Call local Markov AI model
const res = await client.ask('Who created you?');
console.log('Response:', res.text);

2. Multi-Provider Queries (e.g. Google Gemini)

Configuring a specific provider is simple. The SDK handles formatting differences underneath.

import { createKrimsClient } from 'krims-code-sdk';

// Auto-detects process.env.GOOGLE_API_KEY
const client = createKrimsClient({
  provider: 'google', // or 'openai', 'anthropic', 'cohere', 'groq', etc.
  model: 'gemini-2.5-flash'
});

const res = await client.ask('Explain quantum computing in one sentence.');
console.log('Gemini:', res.text);

3. Stateful Chat Sessions

The stateful KrimsSession handles storing message history and appends it automatically to subsequent calls in the session.

import { createKrimsClient } from 'krims-code-sdk';

const client = createKrimsClient({
  provider: 'openai',
  apiKey: 'your-openai-api-key'
});

const session = client.createSession({
  name: 'coding-assistant',
  systemInstruction: 'You are a Senior JavaScript developer.'
});

// Turn 1
let res = await session.ask('What is closures?');
console.log(res.text);

// Turn 2 (Session maintains context of Turn 1 automatically)
res = await session.ask('Can you show me a code example?');
console.log(res.text);

// View cumulative history
console.log(session.getHistory());

API Reference

KrimsClient Options

When calling createKrimsClient(options) or new KrimsClient(options), you can configure:

| Option | Type | Description | Environment Variable Fallback | | :--- | :--- | :--- | :--- | | provider | string | The AI provider to query (e.g. openai, google, anthropic, cohere, groq, together, krims, custom). | - | | model | string | The model identifier. | Default values defined per provider | | apiKey | string | API authorization key. | OPENAI_API_KEY, GOOGLE_API_KEY, ANTHROPIC_API_KEY, etc. | | baseUrl | string | Custom root URL for the endpoints. | KRIMS_BASE_URL (for krims provider only) | | headers | object | Custom headers to append on every request. | - | | timeout | number | Timeout in milliseconds (default: 30000). | - | | fetchFn | function | Custom fetch function (for proxies or custom clients). | globalThis.fetch |

Running Tests

To run the unit tests natively in Node:

npm test