Ai Quiz Generator

File Structure

lib/
├── llm/           → See [references/llm-client.md](references/llm-client.md)
└── quiz/
    ├── types.ts   → See [references/types.md](references/types.md)
    └── tools/     → See [references/function-tools.md](references/function-tools.md)

app/api/quiz/
├── extract-knowledge/
├── generate-quiz/
└── generate-summary/  → See [references/api-routes.md](references/api-routes.md)

Core Workflow

Step 1: Extract Knowledge Points

const response = await fetch("/api/quiz/extract-knowledge", {
  method: "POST",
  body: JSON.stringify({ content: userText }),
});
const { knowledgePoints } = await response.json();
// → [{ id, title, content }, ...]

Step 2: Generate Quiz

const response = await fetch("/api/quiz/generate-quiz", {
  method: "POST",
  body: JSON.stringify({
    points: knowledgePoints,
    character: { name: "孔子", persona: "因材施教...", avatar: "" },
    questionCount: 10,
    quizType: "mixed",
  }),
});
const { questions } = await response.json();
// → [{ id, type, question, options?, answer, explanation }, ...]

Step 3: Generate Summary

const response = await fetch("/api/quiz/generate-summary", {
  method: "POST",
  body: JSON.stringify({
    questions,
    userAnswers: { "q-1": "A", "q-2": "答案文本" },
    character,
  }),
});
const { summary } = await response.json();
// → Markdown 格式的 AI 评语

Key Patterns

Force Structured Output

// Always use tool_choice to force specific function call