Skip to main content

​
Purpose and Responsibilities

  • Parse and summarize the project idea.
  • Extract key features and estimate complexity (low | medium | high) and estimated_agents.
  • Generate 3–5 clarifying questions for the Intent step.

​
Input/Output

​
Input

  • idea: string — Raw idea from the user.
  • sessionId: string — For logging/tracing.

​
Output (IdeaAnalysis in lib/openai.ts)

{
  status: "success";
  stage: "idea_input";
  analysis: {
    summary: string;
    key_features: string[];
    complexity: "low" | "medium" | "high";
    estimated_agents: number;
  };
  clarifying_questions: string[];
  next_stage: "intent_analysis";
}

​
Prompt Engineering

  • System: Instructs the model to act as an “expert idea analysis assistant (IdeaProcessor)” and to respond with valid JSON in the exact structure above. Asks for 3–5 thoughtful clarifying questions.
  • User: Session ID: ${sessionId}\n\nUser Idea: ${idea}
  • Options: response_format: { type: "json_object" }, temperature: 0.7. Model: DEFAULT_MODEL_TYPE (e.g. GROK).
Implementation: processIdea(idea, sessionId, modelType?) in lib/openai.ts.

​
Examples

​
Request (via /api/idea)

{ "idea": "A habit tracker with streaks and reminders." }

​
Response (idea + analysis + questions)

{
  "status": "success",
  "stage": "idea_input",
  "analysis": {
    "summary": "A habit tracker with streaks and reminders.",
    "key_features": ["habit logging", "streaks", "reminders"],
    "complexity": "low",
    "estimated_agents": 3
  },
  "clarifying_questions": [
    "Web, mobile, or both?",
    "Do you need social or sharing features?",
    "Should reminders be push, email, or in-app?"
  ],
  "next_stage": "intent_analysis"
}

​
Customization Options

  • Model: Pass modelType to processIdea (e.g. QWEN, GEMINI).
  • System prompt: Edit the processIdea system string in lib/openai.ts to change tone, structure, or number of questions.
  • JSON schema: If you change IdeaAnalysis, ensure the system prompt and any callers (e.g. /api/idea) stay in sync.