Hologres：AI_GEN_STRUCTURED

模式一：json_schema（推薦）

通過標準 JSON Schema 精確定義輸出結構，模型將嚴格按 schema 產生輸出：

{
    "type": "json_schema",
    "schema": {
        "type": "object",
        "properties": {
            "field_name": {"type": "string"},
            "field_age": {"type": "integer"}
        },
        "required": ["field_name", "field_age"],
        "additionalProperties": false
    }
}

Schema 越精確（類型約束、enum 枚舉、description 描述），輸出品質越高。

模式二：json_object

僅保證輸出為合法 JSON 對象，具體欄位由提示詞控制：

{
    "type": "json_object"
}

使用 json_object 時，需要在 message 提示詞中明確說明期望的 JSON 欄位和格式。

樣本 1：從文本中提取結構化實體

SELECT ai_gen_structured(
    model_name => 'qwen3.7-max',
    message => '從以下文本中提取人員資訊：張三是阿里巴巴的進階工程師，男，漢族，1990年3月15日出生，今年35歲',
    response_format => '{
        "type": "json_schema",
        "schema": {
            "type": "object",
            "properties": {
                "name": {"type": "string"},
                "age": {"type": "integer"},
                "sex": {"type": "string", "enum": ["男", "女"]},
                "birth_day": {"type": ["string", "null"], "format": "date"},
                "民族": {"type": "string", "enum": ["漢族", "蒙古族", "回族", "藏族", "維吾爾族"]}
            },
            "required": ["name", "age", "sex", "birth_day", "民族"],
            "additionalProperties": false
        }
    }'::json
);

輸出：

{"name": "張三", "age": 35, "sex": "男", "birth_day": "1990-03-15", "民族": "漢族"}

樣本 2：json_object 模式

SELECT ai_gen_structured(
    model_name => 'qwen3.7-max',
    message => '請以 JSON 格式返回杭州這座城市的資訊，欄位包括：province（所屬省份）、is_municipality（是否直轄市，布爾值）。',
    response_format => '{"type": "json_object"}'::json
);

輸出：

{"province": "浙江省", "is_municipality": false}

樣本 3：多模態 — 從圖片中提取結構化資訊

SELECT ai_gen_structured(
    model_name => 'qwen3.7-plus',
    message => '識別這張駕駛情境圖片，提取情境結構化資訊。',
    input_file => to_file(
        'oss://your-bucket/path/to/image.jpg',
        'oss-cn-hangzhou-internal.aliyuncs.com',
        'acs:ram::your_account_id:role/your_role'
    ),
    response_format => '{
        "type": "json_schema",
        "schema": {
            "type": "object",
            "properties": {
                "scene": {"type": "string", "description": "情境描述"},
                "time_of_day": {"type": "string", "enum": ["白天", "夜晚", "黃昏"]},
                "weather": {"type": "string"},
                "has_vehicle": {"type": "boolean"}
            },
            "required": ["scene", "time_of_day", "weather", "has_vehicle"],
            "additionalProperties": false
        }
    }'::json
);

樣本 4：批量處理表資料

-- 假設 documents 表有 id 和 content 欄位
SELECT id, ai_gen_structured(
    'qwen3.7-max',
    '提取人員資訊：' || content,
    '{"type":"json_schema","schema":{"type":"object","properties":{"name":{"type":"string"},"age":{"type":"integer"},"city":{"type":"string"}},"required":["name","age","city"],"additionalProperties":false}}'::json,
    'skip',
    '{"temperature":0.1}'::json
) AS extracted_info
FROM documents
ORDER BY id;

批量處理時建議使用 on_error => 'skip'，避免單行異常中斷整個查詢。