Installation

Integrate CLM into your stack in under 2 minutes.

Node.js

The CLM Node.js library is written in TypeScript and includes type definitions.

# Install via npm
npm install @clm/sdk

# Install via yarn
yarn add @clm/sdk

Python

Our Python client supports synchronous and asynchronous execution patterns.

# Install via pip
pip install clm-ai

Authentication

Securely accessing the API.

The CLM API uses API keys for authentication. You can manage your API keys in the dashboard. Your API keys carry many privileges, so be sure to keep them secure.

SDK Usage

import { Client } from '@clm/sdk';

const client = new Client({
  apiKey: process.env.CLM_API_KEY
});

Model Overview

Three distinct architectures for every use case.

CLM currently offers three models. Each is optimized for specific trade-offs between latency, reasoning depth, and cost.

Detailed Breakdown

1. CLM 1.5 (Standard)

The balanced legacy model. Reliable for most standard applications that do not require sub-20ms latency or advanced multi-step reasoning.

2. CLM 2 Flash (Speed)

Built for extreme throughput. CLM 2 Flash is our most cost-effective model, designed for high-volume applications like customer support bots, real-time analytics, and simple data extraction.

3. CLM 2 Pro (Reasoning)

Our flagship "Thinking" model. CLM 2 Pro employs an internal chain-of-thought verification process before outputting tokens. It excels at writing complex software, legal analysis, and scientific research.

Context Window

Understanding memory limitations.

The context window represents the total amount of information (tokens) the model can retain in its "working memory" during a single request-response cycle.

• CLM 1.5: 32,768 tokens (~25k words)
• CLM 2 Flash: 128,000 tokens (~100k words)
• CLM 2 Pro: 200,000 tokens (~150k words)

Chat Completions

POST /v1/chat/completions

Creates a model response for the given chat conversation.

Parameters

• model Required
  E.g., clm-2-pro or clm-2-flash.
• messages Required
  A list of messages comprising the conversation.
• temperature Optional
  0 to 2. Higher values mean more random outputs.

Example: Deep Reasoning (CLM 2 Pro)

const completion = await client.chat.completions.create({
  model: "clm-2-pro",
  messages: [
    { role: "system", content: "You are a senior engineer." },
    { role: "user", content: "Refactor this legacy codebase..." }
  ],
  temperature: 0.2 // Lower temperature for precision
});

Example: High Speed (CLM 2 Flash)

const response = await client.chat.completions.create({
  model: "clm-2-flash",
  messages: [{ role: "user", content: "Summarize this email." }],
  max_tokens: 500
});

Embeddings

POST /v1/embeddings

Get a vector representation of a given input. CLM embeddings are compatible with all major vector databases (Pinecone, Milvus, Chroma).

Example

const embedding = await client.embeddings.create({
  model: "clm-embed-v1",
  input: "The quick brown fox jumped over the lazy dog"
});

console.log(embedding.data[0].embedding);
// Output: [0.00230, -0.0032, ...]

Error Codes

Handling API exceptions.