packages

@loop-engine/adapter-grok

Overview

@loop-engine/adapter-grok wraps xAI's Grok API as a Loop Engine AI actor. Grok uses an OpenAI-compatible API, and this adapter uses the openai npm package pointed at https://api.x.ai/v1. It follows the same governance model as adapter-anthropic and adapter-openai: identical guard enforcement and the same audit trail shape.

Installation

1npm install @loop-engine/adapter-grok openai

Peer dependencies

1openai  ^4.0.0

Basic usage

1import OpenAI from 'openai'
2import { createGrokActorAdapter } from '@loop-engine/adapter-grok'
3 
4"cmt">// Grok uses the OpenAI SDK with xAI's base URL
5const grok = new OpenAI({
6  apiKey: process.env.XAI_API_KEY,
7  baseURL: 'https://api.x.ai/v1',
8})
9 
10const adapter = createGrokActorAdapter(process.env.XAI_API_KEY!, {
11  modelId: 'grok-3',
12  confidenceThreshold: 0.75,
13})
14 
15const { actor, decision } = await adapter.createSubmission({
16  loopId: 'procurement',
17  loopName: 'SCM Procurement',
18  currentState: 'pending_analysis',
19  availableSignals: [
20    {
21      signalId: 'submit_recommendation',
22      name: 'Submit Recommendation',
23      allowedActors: ['ai-agent'],
24    },
25  ],
26  instruction: 'Analyze demand data and recommend a purchase order decision.',
27  evidence: { demandForecast: 0.91, currentStock: 38 },
28})

Configuration reference

| Option | Type | Default | Description | |--------|------|---------|-------------| | modelId | string | grok-3 | Grok model to use (grok-2, grok-2-mini, grok-3, grok-3-mini) | | maxTokens | number | 1024 | Max tokens in response | | systemPrompt | string | — | Optional system prompt prepended to loop context | | confidenceThreshold | number | 0.7 | Minimum confidence required (0-1) | | baseURL | string | https://api.x.ai/v1 | xAI API endpoint (override for testing) |

What the adapter produces

The adapter returns an AIAgentActor with:

• type: "ai-agent"
• provider: "grok"
• modelId — the model used
• confidence — extracted from the model response
• promptHash — SHA-256 of the prompt sent (for audit trail)

Guard enforcement note

The confidence guard runs at the runtime level — not inside the adapter. If you configure a confidence-threshold guard on the transition, the runtime will block the transition if the model's confidence falls below the threshold regardless of what the adapter returns.