---

name: idea-generation description: Work with IdeaForge's AI idea generation system. Triggers: generation flow, AI prompts, scoring system, duplicate detection, real-time logs, generation debugging. Pipeline: API → PromptBuilder → AI → Parse → Dedupe → Save.

Idea Generation

Pipeline

POST /api/generation/generate
  → IdeaGenerationService
    → PromptBuilder (YAML configs)
    → callAI() (Claude/Gemini)
    → parseAIResponse()
    → EmbeddingService (duplicate check)
    → IdeaRepository (save)
    → GenerationLogger (SSE logs)

API

POST /api/generation/generate
{
  "framework": "pain-point",  // optional, random if not specified
  "domain": "Healthcare",      // optional
  "sessionId": "uuid"          // for log tracking
}

AI Response Shape

{
  "name": "Idea Name (max 60 chars)",
  "domain": "Domain → Subdomain",
  "problem": "...",
  "solution": "...",
  "quickSummary": "Elevator pitch",
  "concreteExample": {
    "currentState": "How users handle this today",
    "yourSolution": "How they'd use your product",
    "keyImprovement": "Quantifiable improvement"
  },
  "evaluation": {
    "problemSeverity": { "score": 8, "reasoning": "..." }
  },
  "tags": ["tag1", "tag2"]
}

Scoring

• Per-criterion: 1-10 scale
• Weighted total: 0-100 scale
• Config: evaluation-criteria.yaml

Duplicate Detection

1. Exact: Same domain + problem + solution
2. Semantic: Embedding similarity > 85%

SSE Logs

Stages: INIT → CONFIG_LOAD → PROMPT_BUILD → API_CALL → RESPONSE_PARSE → DUPLICATE_CHECK → DB_SAVE → COMPLETE

Endpoint: GET /api/logs/stream/{sessionId}

Key Files

File	Purpose
services/ideaGenerationService.ts	Orchestration
services/promptBuilder.ts	AI prompts
services/configService.ts	YAML configs
lib/aiProvider.ts	Claude/Gemini