name: Generate YAML from Claude

on:
  workflow_dispatch: 

jobs:
  generate-claude:
    runs-on: ubuntu-latest
    permissions:
      contents: write
      pull-requests: write

    steps:

      # 省略

      - name: Install Python dependencies
        run: |
          pip install -r .github/scripts/requirements_generate_claude_from_pdf.txt

      - name: Generate Claude YAML from PDFs
        env:
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
        run: |
          python .github/scripts/generate_claude_from_pdf.py

      - name: Create Pull Request
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
        run: |
          # Git Configuration
          git config --global user.email "kyash-bot@kyash.co"
          git config --global user.name "kyash-bot"

          # Create new branch name
          BRANCH_NAME="auto-generate-claude-$(date +%Y%m%d-%H%M%S)"

          # Checkout new branch
          git checkout -b "$BRANCH_NAME"

          # Add generated file to git
          git add docs/openapi.yaml
          git commit -m "Generate OpenAPI YAML from PDF specifications

          This YAML file was automatically generated by Claude AI from PDF specifications in docs/spec/.
          
          Important: Please review the generated content carefully as it may contain errors.
          
          Generated with Claude Code"

          # Push branch to origin
          git push origin "$BRANCH_NAME"

          # Create Pull Request
          gh pr create \
            --title "Generate Claude YAML from PDFs API specifications" \
            --body "## Summary
            
            This PR contains an automatically generated OpenAPI YAML file created from PDF specifications in `docs/spec/`.
            
            ## Generated File
            - `docs/openapi.yaml`
            - OpenAPI 3.0.3 specification
            
            ## Source PDFs
            
            The following PDF files were processed:
            $(ls -1 docs/spec/*.pdf | sed 's/^/- /')
            
            ## Important Notes
            
            **This file was generated by AI (Claude)** - Please review carefully:
              
            - [ ] Verify all API endpoints are correct
            - [ ] Check request/response schemas
            - [ ] Validate parameter types and constraints
            - [ ] Confirm field descriptions are accurate
            - [ ] Review error response definitions
            
            ## Next Steps
            
            1. Review the generated YAML file
            2. Make any necessary corrections
            3. Approve and merge if satisfied
            
            Generated with [Claude Code](https://claude.com/claude-code)"

#!/usr/bin/env python3

# 省略

def extract_api_spec_from_pdf(client: anthropic.Anthropic, pdf_path: str, pdf_name: str) -> Dict[str, Any]:
    prompt = """このPDFファイルは日本語で書かれたAPI仕様書です。
このAPI仕様書を分析して、OpenAPI 3.0.3形式のYAML仕様の一部として出力してください。

以下の情報を抽出してください：
1. APIエンドポイント（パス）
2. HTTPメソッド（GET, POST, PUT, DELETEなど）
3. リクエストパラメータ（クエリ、パス、ボディ）
4. リクエストボディのスキーマ（該当する場合）
5. レスポンスの形式とスキーマ
6. 各フィールドの説明、型、制約（必須/任意、最大長など）
7. エラーレスポンスの定義

出力は以下の形式のJSONで返してください：

# 省略

重要な注意事項：
- 日本語の説明はそのまま維持してください
- データ型は適切なOpenAPI型（string, integer, number, boolean, array, objectなど）に変換してください
- 数値型の場合、整数はinteger、小数はnumberを使用してください
- 必須/任意の情報は required 配列に反映してください
- レスポンス例がある場合は examples に含めてください
- エラーレスポンスも responses に含めてください（400, 401, 500など）
- 出力は有効なJSONのみで、説明文は含めないでください
"""

    print(f"  Sending request to Claude API...")
    message = client.messages.create(
        model="claude-sonnet-4-5-20250929",
        max_tokens=8192,
        messages=[
            {
                "role": "user",
                "content": [
                    {
                        "type": "document",
                        "source": {
                            "type": "base64",
                            "media_type": "application/pdf",
                            "data": pdf_data,
                        },
                    },
                    {
                        "type": "text",
                        "text": prompt
                    }
                ],
            }
        ],
    )

    # レスポンスからテキストを抽出
    response_text = message.content[0].text
    print(f"  Received response from Claude API")

    try:
        spec = json.loads(response_text)
        print(f"  ✓ Successfully parsed API spec")
        return spec
    except json.JSONDecodeError as e:
        print(f"  ✗ Error parsing JSON: {e}")
        print(f"  Response preview: {response_text[:300]}...")
        return None

def main():
    print("=" * 60)
    print("PDF to OpenAPI YAML Generator")
    print("=" * 60)

    # 環境変数からAPIキーを取得
    api_key = os.getenv('ANTHROPIC_API_KEY')
    if not api_key:
        print("❌ Error: ANTHROPIC_API_KEY environment variable is not set")
        sys.exit(1)

    # Claude APIクライアントを初期化
    client = anthropic.Anthropic(api_key=api_key)
    print("✓ Claude API client initialized\n")

    # PDFファイルのディレクトリ
    pdf_dir = Path('docs/spec')
    if not pdf_dir.exists():
        print(f"❌ Error: Directory {pdf_dir} does not exist")
        sys.exit(1)

    # PDFファイルを取得（.DS_Storeなどを除外）
    pdf_files = sorted([f for f in pdf_dir.glob('*.pdf') if not f.name.startswith('.')])
    print(f"Found {len(pdf_files)} PDF file(s) in {pdf_dir}\n")

    if not pdf_files:
        print("❌ Error: No PDF files found")
        sys.exit(1)

    # 各PDFからAPI仕様を抽出
    api_specs = []
    for i, pdf_file in enumerate(pdf_files, 1):
        print(f"[{i}/{len(pdf_files)}] Processing: {pdf_file.name}")
        try:
            spec = extract_api_spec_from_pdf(client, str(pdf_file), pdf_file.name)
            if spec:
                api_specs.append(spec)
                print(f"  ✓ Successfully extracted spec from {pdf_file.name}\n")
            else:
                print(f"  ⚠ Failed to extract spec from {pdf_file.name}\n")
        except Exception as e:
            print(f"  ❌ Error processing {pdf_file.name}: {e}\n")
            continue

    if not api_specs:
        print("❌ Error: No API specifications were extracted")
        sys.exit(1)

    print("=" * 60)
    print(f"✓ Successfully extracted {len(api_specs)} API specification(s)")
    print("=" * 60)

    # OpenAPI仕様にマージ
    print("\nMerging API specifications into OpenAPI format...")
    openapi_spec = merge_api_specs_to_openapi(api_specs)
    print(f"✓ Merged into single OpenAPI specification")
    print(f"  Total paths: {len(openapi_spec['paths'])}")

    # YAMLファイルとして出力
    output_file = Path('docs/openapi.yaml')
    output_file.parent.mkdir(parents=True, exist_ok=True)

    print(f"\nWriting to {output_file}...")
    with open(output_file, 'w', encoding='utf-8') as f:
        yaml.dump(openapi_spec, f, allow_unicode=True, sort_keys=False, default_flow_style=False)

    print("=" * 60)
    print(f"✅ SUCCESS: OpenAPI specification written to {output_file}")
    print("=" * 60)
    print("\n⚠️  重要: 生成されたYAMLファイルは必ずレビューしてください")
    print("   AIが生成した内容のため、誤りがある可能性があります\n")

if __name__ == '__main__':
    main()