Briefing Generator

Dynamically generate contextual follow-up questions to extract the information the panel needs for a high-quality deliberation.

Overview

The Briefing Generator takes a user's decision question and produces 5-7 tailored multiple-choice questions. The answers are compiled into a context string that's injected into the deliberation prompt — giving the panel the constraints, priorities, and domain context it needs.

Note

The questions are AI-generated, not hardcoded. They adapt to the domain and specifics of the user's question. A question about hiring produces different follow-ups than one about technical architecture.

Usage

import { BriefingGenerator } from 'decisionmemos';
import { createDecisionEngine } from 'decisionmemos';

const engine = createDecisionEngine();
const briefing = new BriefingGenerator();

// 1. Generate questions from the user's question
const questions = await briefing.generateQuestions(
  "Should we hire a CTO or promote internally?",
  engine.getClient('openai') // Uses one model to generate questions
);

// questions = [
//   { id: "q1", prompt: "How large is your engineering team?",
//     options: [...], allowMultiple: false },
//   { id: "q2", prompt: "What's the primary gap you're trying to fill?",
//     options: [...], allowMultiple: true },
//   ...
// ]

// 2. Collect answers (e.g. from a UI or programmatically)
const answers = {
  q1: ["medium"],      // 10-50 people
  q2: ["strategy", "fundraising"],
  q3: ["skip"],        // User skipped this question
};

// 3. Compile into context string
const context = briefing.compileContext(questions, answers);
// → "How large is your engineering team? 10-50 people.
//    What's the primary gap? Strategy, Fundraising."

// 4. Use the context in the deliberation
// (context is automatically injected into the prompt)

Question schema

interface BriefingQuestion {
  id: string;
  prompt: string;           // Short question (< 15 words)
  options: Array<{
    id: string;
    label: string;          // Short label (< 10 words)
  }>;
  allowMultiple: boolean;   // Can select multiple options?
}

Fallback behaviour

If the AI generation fails (network error, malformed response, etc.), the Briefing Generator falls back to a set of generic questions about timeline, scale, budget, and priorities. These are less tailored but still extract useful context.