@pmcollab/coworkstream-engine-openai

Coding agents: read packages/workstream-engine/AGENTS.md for the engine integration recipe before wiring this adapter. Migrating an existing setup? See packages/workstream-engine/MIGRATION.md.

OpenAI AgentRunner adapter for @pmcollab/coworkstream-engine. Uses chat completions with JSON Schema response_format for structured output.

Install

npm install @pmcollab/coworkstream-engine-openai openai

Use

import OpenAI from 'openai'
import { createEngine } from '@pmcollab/coworkstream-engine'
import { createOpenAIAgent } from '@pmcollab/coworkstream-engine-openai'

const client = new OpenAI({ apiKey: process.env.OPENAI_API_KEY })

const agent = createOpenAIAgent({
  id: 'risk',
  client,
  model: 'gpt-4o',
  instructions: 'You are a risk reviewer. Evaluate proposed actions for risk to revenue.',
})

The model is asked to return a JSON object matching:

{ position: string, reasoning: string, confidence: number }

Position is taken straight from the JSON. Malformed responses produce a { confidence: 0, error } result so the engine can de-weight that agent without crashing.

Options

createOpenAIAgent({
  id, client,                    // required
  name?, avatar?, capabilities?,
  model?: string,                // default 'gpt-4o'
  instructions?: string,
  temperature?: number,          // default 0.2
  defaultConfidence?: number,    // default 0.7
  formatItem?: (item, ctx) => string,
})

License

Commercial. See LICENSE in the repository root.