@barric/openai

barric firewall wrapper for the OpenAI SDK. Drop-in replacement that scans every chat.completions.create call for prompt injection, PII leaks, and other threats.

Install

npm install @barric/core @barric/openai openai

Quick Start

import OpenAI from 'openai'
import { createFirewall } from '@barric/core'
import { barricOpenAI } from '@barric/openai'

const firewall = createFirewall({
  rules: ['prompt-injection', 'pii-redaction'],
  injection: { threshold: 0.7, action: 'block' },
  pii: { mode: 'redact', types: ['email', 'phone', 'ssn'] },
})

const client = barricOpenAI(new OpenAI(), firewall)

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

console.log(response.choices[0].message.content)
console.log(response._barric.events) // firewall scan results

How It Works

barricOpenAI returns a drop-in replacement for the OpenAI client:

1. Extracts the last user message from params.messages
2. Runs it through inbound scanners (injection detection, PII redaction, rate limiting, encoding detection, input limits)
3. If the input passes, the sanitized message replaces the original and the real OpenAI API call executes
4. Outbound scanners run on the response (output limits, system prompt leak detection)
5. The response is returned with a _barric property containing firewall metadata

System and assistant messages pass through untouched.

Handling Blocked Requests

When the firewall blocks a request, an error is thrown before the API call:

try {
  const response = await client.chat.completions.create({
    model: 'gpt-4o',
    messages: [{ role: 'user', content: maliciousInput }],
  })
} catch (err) {
  if (err._barric?.blocked) {
    console.log('Blocked by:', err._barric.events[0].scanners)
  } else {
    throw err // Regular OpenAI error
  }
}

Per-request Context

const client = barricOpenAI(new OpenAI(), firewall, {
  context: { userId: 'user-123', ip: '192.168.1.1' },
})

License

MIT