assert response.json() == snapshot(
        {
            "components": {
                "schemas": {
                    "Address": {
                        # NOTE: the description of this model shows only the public-facing text, before the `\f` in docstring
                        "description": "This is a public description of an Address\n",
                        "properties": {

And then you can set headers in that *temporal* response object.

{* ../../docs_src/response_headers/tutorial002_py310.py hl[1, 7:8] *}

And then you can return any object you need, as you normally would (a `dict`, a database model, etc).

And if you declared a `response_model`, it will still be used to filter and convert the object you returned.

* **No serialization**: Just pass the raw, unstructured data around. Serialization is for people who don't trust their LLMs.
* **Embrace modern AI coding practices**: Leave everything up to an LLM to decide. The model knows best. Always.

import java.nio.file.Files;
import java.nio.file.Path;
import java.util.ArrayList;
import java.util.List;
import java.util.ListIterator;
import java.util.Objects;

import org.apache.maven.api.model.InputLocation;
import org.apache.maven.api.model.Plugin;
import org.eclipse.aether.AbstractRepositoryListener;
import org.eclipse.aether.RepositoryEvent;
import org.eclipse.aether.RepositorySystemSession;
import org.eclipse.aether.RequestTrace;

            "publishPluginsToTestRepository",
        ) -> false

        // Core tasks
        task.name in listOf(
            "components",
            "dependantComponents",
            "model",
        ) -> true

        task.name.startsWithAnyOf(
            "publish",
            "idea",
        ) -> true

        task is GradleBuild -> true

        // gradle/gradle build tasks

首先，确保你创建并激活了[虚拟环境](../virtual-environments.md)，然后安装 `sqlmodel`：

<div class="termy">

```console
$ pip install sqlmodel
---> 100%
```

</div>

## 创建含有单一模型的应用 { #create-the-app-with-a-single-model }

我们先创建应用的最简单的第一个版本，只有一个 **SQLModel** 模型。

稍后我们将通过下面的**多个模型**提高其安全性和多功能性。🤓

### 创建模型 { #create-models }

导入 `SQLModel` 并创建一个数据库模型：

首先，你需要匯入它：

{* ../../docs_src/body_fields/tutorial001_an_py310.py hl[4] *}


/// warning

請注意，`Field` 是直接從 `pydantic` 匯入的，不像其他（如 `Query`、`Path`、`Body` 等）是從 `fastapi` 匯入。

///

## 宣告模型屬性 { #declare-model-attributes }

接著你可以在模型屬性上使用 `Field`：

{* ../../docs_src/body_fields/tutorial001_an_py310.py hl[11:14] *}

`Field` 的用法與 `Query`、`Path`、`Body` 相同，擁有相同的參數等。

/// note | 技術細節

	//  a large number of policies
	r, err := ies.client.Get(ctx, iamConfigPoliciesPrefix, etcd.WithPrefix())
	if err != nil {
		return err
	}

	// Parse all values to construct the policies data model.
	for _, kvs := range r.Kvs {
		if err = ies.getPolicyDocKV(ctx, kvs, m); err != nil && !errors.Is(err, errNoSuchPolicy) {
			return err
		}
	}
	return nil
}

 * For 4.0.0 models: applies limited inference (parent-related only).
 * For 4.1.0+ models: applies full inference optimizations.
 * Removes redundant information that can be inferred by Maven during model building.
 */
@Named
@Singleton
@Priority(30)
public class InferenceStrategy extends AbstractUpgradeStrategy {

    @Override
    public boolean isApplicable(UpgradeContext context) {

    public static final String ERROR_TIMEOUT = "timeout";

    /** Error code for context length exceeded. */
    public static final String ERROR_CONTEXT_LENGTH_EXCEEDED = "context_length_exceeded";

    /** Error code for model not found. */
    public static final String ERROR_MODEL_NOT_FOUND = "model_not_found";

    /** Error code for invalid response from the LLM provider. */
    public static final String ERROR_INVALID_RESPONSE = "invalid_response";