Skip to main content

Documentation Index

Fetch the complete documentation index at: https://cognisafeltd.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

​
Installation

npm install cognisafe
The package is published to npm as cognisafe. It ships with full TypeScript types — no @types/ package needed.

​
TypeScript support

The SDK is written in TypeScript and exports proper type definitions. It works with both import (ESM) and require (CJS): 
// ESM / TypeScript
import cognisafe from 'cognisafe';

// CommonJS
const cognisafe = require('cognisafe');

​
Configuration

Call cognisafe.configure() once at startup, before any LLM calls: 
import cognisafe from 'cognisafe';

cognisafe.configure({
  apiKey: 'csk_your_key_here',    // or reads COGNISAFE_API_KEY
  projectId: 'my-app',            // or reads COGNISAFE_PROJECT_ID
  proxyUrl: 'http://localhost:8080',  // default; omit for cloud
  apiUrl: 'http://localhost:8000',    // default; omit for cloud
});

​
Configuration options

OptionTypeRequiredDefaultDescription
apiKeystringYes*COGNISAFE_API_KEYYour project API key (csk_...)
projectIdstringYes*COGNISAFE_PROJECT_IDProject identifier for grouping requests
proxyUrlstringNohttp://localhost:8080Cognisafe Go proxy URL
apiUrlstringNohttp://localhost:8000Cognisafe FastAPI backend URL
* Required if the corresponding environment variable is not set.

​
Patching providers

​
OpenAI

import cognisafe from 'cognisafe';
import OpenAI from 'openai';

cognisafe.configure({ apiKey: 'csk_...', projectId: 'my-app' });
cognisafe.patchOpenAI();

const client = new OpenAI(); // baseURL is now the Cognisafe proxy

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: 'Hello, world!' }],
});

console.log(response.choices[0].message.content);
patchOpenAI() sets the baseURL on the default OpenAI client to the Cognisafe proxy. The proxy forwards to https://api.openai.com and logs every call asynchronously.

​
Anthropic

import cognisafe from 'cognisafe';
import Anthropic from '@anthropic-ai/sdk';

cognisafe.configure({ apiKey: 'csk_...', projectId: 'my-app' });
cognisafe.patchAnthropic();

const client = new Anthropic();

const message = await client.messages.create({
  model: 'claude-opus-4-5',
  max_tokens: 1024,
  messages: [{ role: 'user', content: 'Hello.' }],
});

console.log(message.content[0].text);
Anthropic uses direct mode — the SDK wraps messages.create and ships payloads to the backend after the response is returned to your code.

​
The trace wrapper

For custom LLM endpoints or any call not covered by a patch method: 
import cognisafe from 'cognisafe';

const callMyModel = cognisafe.trace(
  async (prompt: string): Promise<string> => {
    const response = await fetch('https://internal.example.com/v1/chat', {
      method: 'POST',
      headers: { 'Content-Type': 'application/json' },
      body: JSON.stringify({ messages: [{ role: 'user', content: prompt }] }),
    });
    const data = await response.json();
    return data.choices[0].message.content;
  },
  { model: 'my-internal-model' }
);

const result = await callMyModel('Summarise this text.');
trace returns a wrapped version of your function. The wrapper captures inputs and outputs and sends them to /internal/log after the function resolves — your code is not blocked.
The Node SDK is published under the same package name (cognisafe) as the Python SDK but is a completely separate implementation. Install via npm install cognisafe, not pip.