@kellanjs/actioncraft/README.md

Fluent, type-safe builder for Next.js server actions.

⚠️🚧 The library hasn't reached a stable release yet. Expect bugs and potentially breaking API changes until then.

# ActionCraft

Streamline your server actions.

- **🔒 Full Type Safety** - End-to-end TypeScript inference from input to output
- **📝 Schema Validation** - Works with Zod and any Standard Schema V1 compliant library
- **🎯 Fluent API** - Readable, discoverable builder pattern
- **⚡ Progressive Enhancement** - Works with and without JavaScript enabled
- **🔄 React Integration** - Built-in support for `useActionState` and form handling
- **🛡️ Error Management** - Structured error handling with custom error types
- **🔗 Lifecycle Hooks** - Callbacks for start, success, error, and completion events
- **📋 Form State Preservation** - Automatic form value retention on validation errors

## Table of Contents

- [Quick Start](#quick-start)
  - [Installation](#installation)
  - [Overview](#overview)
  - [Example](#example)
  - [Result Format](#result-format)
- [Walkthrough](#walkthrough)
  - [.create() - Configure Your Action](#create)
  - [.schemas() - Add Validation](#schemas)
  - [.errors() - Define Custom Errors](#errors)
  - [.action() - Implement Business Logic](#action)
  - [.callbacks() - Add Lifecycle Hooks](#callbacks)
  - [.craft() - Build Your Action](#craft)
- [Using Your Actions](#using-your-actions)
  - [Basic Usage](#basic-usage)
  - [Error Handling](#error-handling)
  - [React Forms with useActionState](#react-forms-with-useactionstate)
  - [Progressive Enhancement](#progressive-enhancement)
- [Complete Example](#complete-example)
- [Integrations](#integrations)
  - [Utilities](#utilities)
  - [React Query](#react-query)
- [Advanced Features](#advanced-features)
  - [Bind Arguments](#bind-arguments)
- [Type Inference Utilities](#type-inference-utilities)

## Quick Start

### Installation

```sh
npm install @kellanjs/actioncraft
```

### Overview

ActionCraft makes it easy to create type-safe server actions with first-class error-handling support. Here's the basic structure you'll follow:

```typescript
const action = create(...)
  .schemas(...)
  .errors(...)
  .action(...)
  .callbacks(...)
  .craft();
```

ActionCraft uses a fluent builder pattern, making it simple to chain one method after the next to create a full-featured server action. The order in which these methods are defined is important for type inference to work properly, so you'll see this same structure repeated often throughout the documentation. Always make sure to chain your methods together like this for the best experience!

### Example

With this basic structure in mind, let's see what a more detailed example looks like:

```typescript
"use server";

import { create } from "@kellanjs/actioncraft";
import { z } from "zod";

const newUserInputSchema = z.object({
  name: z.string(),
  email: z.string().email(),
  age: z.number(),
});

export const createNewUser = create()
  // Define the validation schema
  .schemas({
    inputSchema: newUserInputSchema,
  })
  // Define any errors that can occur in your action
  .errors({
    unauthorized: () =>
      ({
        type: "UNAUTHORIZED",
        message: "You don't have permission to create users",
      }) as const,
    emailTaken: (email: string) =>
      ({
        type: "EMAIL_TAKEN",
        message: `The email "${email}" is already registered`,
        email,
      }) as const,
  })
  // Define your server action logic
  .action(async ({ input, errors }) => {
    // These are your validated input values
    const { name, email, age } = input;

    // If an error occurs, just return the result of the appropriate error function
    if (!hasPermission()) return errors.unauthorized();

    if (await emailExists(email)) return errors.emailTaken(email);

    // Additional business logic here...

    return { newUser };
  })
  // Define lifecycle callbacks (optional)
  .callbacks({
    onSettled: (result) => {
      // Log what happened if you want
    },
  })
  // Finally, build the full type-safe action
  .craft();
```

### Result Format

Server actions work best when you're returning serializable data. Throwing errors is less effective in this context, because Next.js will sanitize Error class objects that are thrown in your action, leaving you without useful error information on the client. You might see something in development, but in production, if you try to display `error.message`, you'll likely see something along the lines of: "An error occurred in the Server Components render. The specific message is omitted in production builds to avoid leaking sensitive details."

ActionCraft was designed with this fundamental behavior in mind! Instead of throwing errors, we're working exclusively with structured, serializable objects every step of the way, so errors in your action will always return the data you need.

The default result format should feel pretty familiar: `{ success: true, data: T } | { success: false, error: E }`

We'll look at errors in more detail later. But here's a simple example of one way you might work with an action result on the client:

```typescript
const handleCreateNewUser = async (userData) => {
  // Call your server action like you normally would and get the result
  const result = await createNewUser(userData);

  if (result.success) {
    // If the action was successful, then you get back typed return data
    toast.success("User created:", result.data.newUser);
  } else {
    // If the action was unsuccessful, then you get type-safe error handling
    switch (result.error.type) {
      case "INPUT_VALIDATION":
        handleInputValidationErrorLogic();
        break;
      case "EMAIL_TAKEN":
        showError(`Email ${result.error.email} is already taken`);
        break;
      case "UNAUTHORIZED":
        handleAuthErrorLogic();
        break;
      case "UNHANDLED":
        handleUncaughtExceptions();
        break;
    }
  }
};
```

## Walkthrough

Now that we've covered the basic structure of an action and looked at a simple example, let's take a more detailed look at each method in the chain and what it can do.

### .create()

Actions always begin with the `create` method. You can use it with no arguments for sensible defaults, or pass a configuration object to customize behavior:

```typescript
// We're just taking the default configuration here
const action = create()
  .schemas(...)
  .errors(...)
  .action(...)
  .callbacks(...)
  .craft();
```

#### Configuration Options

When you want to customize behavior, pass a configuration object:

```typescript
const action = create({
    useActionState: true,
    resultFormat: "api",
    validationErrorFormat: "flattened",
    handleThrownError: (error) => ({
      type: "CUSTOM_ERROR",
      message: error.message,
    }) as const,
  })
  .schemas(...)
  .errors(...)
  .action(...)
  .callbacks(...)
  .craft();
```

##### `useActionState: boolean`

**Default:** `false`

Set to `true` to make your action compatible with React's `useActionState` hook:

```typescript
const action = create({ useActionState: true })
  .schemas(...)
  .errors(...)
  .action(...)
  .callbacks(...)
  .craft();

// Now you can use it with useActionState like this:
const [state, formAction] = useActionState(action, initial(action));
```

##### `resultFormat: "api" | "functional"`

**Default:** `"api"`

ActionCraft supports two different return formats:

- **`"api"`**: `{ success: true, data: T } | { success: false, error: E }`
- **`"functional"`**: `{ type: "ok", value: T } | { type: "err", error: E }`

##### `validationErrorFormat: "flattened" | "nested"`

**Default:** `"flattened"`

Controls how validation errors are structured:

- **`"flattened"`**: Returns a flat array of error messages
- **`"nested"`**: Returns a nested object matching your schema structure

##### `handleThrownError: (error: unknown) => UserDefinedError`

By default, ActionCraft catches thrown errors and returns a structured error with type `"UNHANDLED"`. You can customize this behavior by passing an error handler function of your own:

```typescript
const action = create({
  handleThrownError: (error) =>
    ({
      type: "CUSTOM_ERROR",
      message: error instanceof Error ? error.message : "Something went wrong",
      timestamp: new Date().toISOString(),
    }) as const,
});
```

You can even implement more complex logic if you want:

```typescript
handleThrownError: (error: unknown) => {
  if (error instanceof Error) {
    if (error.message.includes("ECONNREFUSED")) {
      return {
        type: "NETWORK_ERROR",
        message: "Unable to connect to external service",
        originalError: error.message,
      } as const;
    }

    if (error.message.includes("timeout")) {
      return {
        type: "TIMEOUT_ERROR",
        message: "Operation timed out",
        originalError: error.message,
      } as const;
    }

    if (error.message.includes("unauthorized")) {
      return {
        type: "AUTHENTICATION_ERROR",
        message: "Authentication failed",
        originalError: error.message,
      } as const;
    }

    // Generic error transformation
    return {
      type: "CUSTOM_HANDLED_ERROR",
      message: `Custom handler caught: ${error.message}`,
      originalError: error.message,
    } as const;
  }

  // Handle non-Error objects
  return {
    type: "UNKNOWN_ERROR_TYPE",
    message: "An unknown error occurred",
    originalError: String(error),
  } as const;
};
```

ActionCraft's types are smart enough to infer all of these possibilities back on the client:

```typescript
if (!result.success) {
  console.log(result.error.type);
  // type: "INPUT_VALIDATION" | "INITIAL_STATE" | "NETWORK_ERROR" | "TIMEOUT_ERROR" | "AUTHENTICATION_ERROR" | "CUSTOM_HANDLED_ERROR" | "UNKNOWN_ERROR_TYPE"
}
```

Pretty cool!

### .schemas()

With our action configured, let's add validation using schemas. ActionCraft supports any library that implements the **Standard Schema V1** interface. Validation is handled automatically - you just need to provide the schemas:

```typescript
const action = create()
  .schemas({
    inputSchema,
    outputSchema,
    bindSchemas,
  })
  .errors(...)
  .action(...)
  .callbacks(...)
  .craft();
```

#### Schema Options

##### `inputSchema?: StandardSchemaV1`

Validates user input passed to the action. If validation fails, an "INPUT_VALIDATION" error is returned to the client.

##### `outputSchema?: StandardSchemaV1`

Validates the data returned from your action. If validation fails, an "OUTPUT_VALIDATION" error is passed to callbacks, but the client always receives an "UNHANDLED" error (this is not affected by `handleThrownError`).

##### `bindSchemas?: StandardSchemaV1[]`

Validates arguments bound to the action with `.bind()`. If validation fails, a "BIND_ARGS_VALIDATION" error is returned to the client.

### .errors()

Now that we have validation set up, let's define custom errors that our action can return. ActionCraft makes error handling really easy by letting you define structured error types:

```typescript
const action = create()
  .schemas(...)
  .errors({
    unauthorized: () =>
      ({
        type: "UNAUTHORIZED",
        message: "You don't have permission to perform this action",
      }) as const,
    notFound: (id: string) =>
      ({
        type: "NOT_FOUND",
        message: `User with ID ${id} not found`,
        id,
      }) as const,
    emailTaken: (email: string) =>
      ({
        type: "EMAIL_TAKEN",
        message: `The email "${email}" is already registered`,
        email,
      }) as const,
  })
  .action(...)
  .callbacks(...)
  .craft();
```

#### Error Structure

Each error is defined as a function called an **ErrorDefinition**:

- **Takes any arguments** you want (like IDs, emails, etc.)
- **Returns a UserDefinedError** object with:
  - `type`: A string discriminator (required)
  - `message`: Human-readable error message (optional)
  - Any other custom fields you want

#### Why the `as const` Assertion?

The `as const` assertion is **required** for proper TypeScript inference. It ensures your error types are treated as literal types rather than generic:

```typescript
// ❌ Without 'as const' - TypeScript infers { type: string, message: string }
badErrorDefinition: () => ({ type: "ERROR", message: "Something went wrong" });

// ✅ With 'as const' - TypeScript infers { type: "ERROR", message: "Something went wrong" }
goodErrorDefinition: () =>
  ({ type: "ERROR", message: "Something went wrong" }) as const;
```

#### Reusing Common Errors

Since error definitions are just functions, you can easily share common errors between actions:

```typescript
// common-errors.ts
export const unauthorized = () =>
  ({
    type: "UNAUTHORIZED",
    message: "You don't have permission to perform this action",
  }) as const;

export const rateLimited = () =>
  ({
    type: "RATE_LIMITED",
    message: "Too many requests. Please try again later.",
  }) as const;

export const notFound = (resource: string, id: string) =>
  ({
    type: "NOT_FOUND",
    message: `${resource} with ID ${id} not found`,
    resource,
    id,
  }) as const;
```

```typescript
// my-action.ts
export const action = create()
  .schemas(...)
  .errors({
    // Easily use common shared errors
    unauthorized,
    rateLimited,
    notFound,
    // Plus any action-specific errors
    emailTaken: (email: string) =>
      ({ type: "EMAIL_TAKEN", email }) as const,
  })
  .action(...)
  .callbacks(...)
  .craft();
```

#### Using Errors in Your Action

Once defined, you can use these errors in your action logic. When an error occurs, just call and return that particular error function:

```typescript
const action = create()
  .schemas(...)
  .errors(...)
  .action(async ({ input, errors }) => {
    // Check permissions
    if (!hasPermission(input.userId)) {
      return errors.unauthorized();
    }

    // Find user
    const user = await findUser(input.userId);
    if (!user) {
      return errors.notFound(input.userId);
    }

    // Success case
    return { user };
  })
  .callbacks(...)
  .craft();
```

### .action()

The `action` method is where you implement the core functionality of your server action. ActionCraft provides several helpful parameters to make things quick and easy:

```typescript
const action = create()
  .schemas(...)
  .errors(...)
  .action(async ({ input, bindArgs, errors, metadata }) => {
    // Action logic here
  })
  .callbacks(...)
  .craft();
```

#### Action Parameters

##### `input`

Contains the validated input values (or `undefined` if no input schema was provided).

##### `bindArgs`

Contains an array of validated bound argument values (or an empty array if no bind schemas were provided).

##### `errors`

Contains all the ErrorDefinition functions you defined in the `.errors()` method.

##### `metadata`

Contains additional request information:

- `rawInput`: The original, unvalidated input data
- `rawBindArgs`: The original, unvalidated bound arguments array
- `prevState`: Previous state (when using `useActionState`)

### .callbacks()

Sometimes you need to hook into the action lifecycle for logging, analytics, or other side effects. The `callbacks` method lets you define functions that run at key moments:

```typescript
const action = create()
  .schemas(...)
  .errors(...)
  .action(...)
  .callbacks({
    onStart: ({metadata}) => { ... },
    onSuccess: ({data}) => { ... },
    onError: ({error}) => { ... },
    onSettled: ({ result }) => { ... },
  })
  .craft();
```

#### Callback Types

##### `onStart?: (params: { metadata }) => Promise<void> | void`

Executes first, before any validation or action logic has occurred.

##### `onSuccess?: (params: { data, metadata }) => Promise<void> | void`

Executes when your action completes successfully. The `data` parameter contains your action's return value.

##### `onError?: (params: { error, metadata }) => Promise<void> | void`

Executes when your action returns an error (custom errors, validation failures, or unhandled exceptions).

##### `onSettled?: (params: { result, metadata }) => Promise<void> | void`

Executes after your action completes, regardless of success or failure. Useful for cleanup or logging.

Note: All callback methods support async operations and won't affect your action's result, even if they throw errors.

### .craft()

Every action should end with the `craft` method. It doesn't take any arguments or anything like that. Its only purpose is to build the final action based on everything that has been defined in the previous methods.

```typescript
const action = create()
  .schemas(...)
  .errors(...)
  .action(...)
  .callbacks(...)
  .craft(); // Returns your fully-typed server action
```

This is the final step in the builder chain. Once you call `craft()`, you have a complete, type-safe server action ready to export and use in your application.

## Using Your Actions

Now that you know how to build actions with ActionCraft, let's see how you can use them in your application.

### Basic Usage

You can call your action like any async function:

```typescript
const result = await createNewUser({
  name: "John",
  email: "john@example.com",
  age: 25,
});

if (result.success) {
  // Action succeeded
  console.log("User created:", result.data.newUser);
} else {
  // Action failed
  console.log("Error:", result.error.type);
  console.log("Message:", result.error.message);
}
```

### Error Handling

Thanks to some carefully crafted types, you can always determine exactly what kind of error you're dealing with:

```typescript
const result = await createNewUser(formData);

if (!result.success) {
  switch (result.error.type) {
    case "INPUT_VALIDATION":
      showValidationErrors(result.error.issues);
      break;
    case "UNAUTHORIZED":
      redirectToLogin();
      break;
    case "EMAIL_TAKEN":
      showError(`Email ${result.error.email} is already taken`);
      break;
    case "UNHANDLED":
      showGenericError();
      break;
  }
}
```

### React Forms with useActionState

For React forms, you can use actions configured for `useActionState`:

```typescript
const updateUser = create({ useActionState: true })
  .schemas(...)
  .errors(...)
  .action(...)
  .callbacks(...)
  .craft();
```

When `useActionState: true` is set, your action's return type changes to include a `values` field. This field contains the raw input values that were last passed to the action. However, on successful executions where an input schema is defined, it contains the validated input values instead.

#### The `initial()` Helper

When using `useActionState`, you have to provide the hook with a proper initial state that matches the return type of your action. That's where ActionCraft's `initial` function comes in. It returns a special error object with type `"INITIAL_STATE"` that you can use to detect when the form hasn't been submitted yet:

```typescript
function UserForm() {
  const [state, action] = useActionState(updateUser, initial(updateUser));
  // `state` initializes as:
  // { success: false,
  //   error: { type: "INITIAL_STATE", message: "Action has not been executed yet" },
  //   values: undefined }

  return (
    <form action={action}>
      <input name="name" defaultValue={state.values?.name} />
      <input name="email" defaultValue={state.values?.email} />

      {!state.success && state.error.type !== "INITIAL_STATE" && (
        <p>Error: {state.error.message}</p>
      )}

      <button type="submit">Update User</button>
    </form>
  );
}
```

### Progressive Enhancement

By providing a schema which supports FormData, your action can work with or without JavaScript. For example, when using Zod, you can use the `zod-form-data` library to provide FormData support for your action:

```typescript
// This action handles FormData from server-side form submissions
const serverAction = create({ useActionState: true })
  .schemas({
    inputSchema: zfd.formData({
      name: zfd.text(),
      email: zfd.text(z.string().email()),
    }),
  })
  .action(async ({ input }) => {
    // Save the validated user data to database
    const user = await db.user.create({
      data: {
        name: input.name,
        email: input.email,
      },
    });

    // Send welcome email
    await sendWelcomeEmail(user.email);

    return { user };
  })
  .craft();
```

## Complete Example

Now that we've gone over how to create actions and how to use them on the client, let's check out an example that puts a lot of these ideas together:

```typescript
"use server";

import { create } from "@kellanjs/actioncraft";
import { revalidatePath } from "next/cache";
import { z } from "zod";

const updateProfileSchema = z.object({
  name: z.string().min(1, "Name is required"),
  email: z.string().email("Invalid email"),
  bio: z.string().max(500, "Bio must be under 500 characters"),
});

export const updateProfile = create({ useActionState: true })
  .schemas({ inputSchema: updateProfileSchema })
  .errors({
    unauthorized: () =>
      ({ type: "UNAUTHORIZED", message: "Please log in" }) as const,
    emailTaken: (email: string) =>
      ({
        type: "EMAIL_TAKEN",
        message: `Email ${email} is already taken`,
        email,
      }) as const,
    rateLimited: () =>
      ({
        type: "RATE_LIMITED",
        message: "Too many requests. Please try again later.",
      }) as const,
  })
  .action(async ({ input, errors }) => {
    // Check authentication
    const session = await getSession();
    if (!session) return errors.unauthorized();

    // Check rate limiting
    if (await isRateLimited(session.userId)) {
      return errors.rateLimited();
    }

    // Check if email is taken
    const existingUser = await getUserByEmail(input.email);
    if (existingUser && existingUser.id !== session.userId) {
      return errors.emailTaken(input.email);
    }

    // Update user
    const updatedUser = await updateUser(session.userId, input);

    return { user: updatedUser };
  })
  .callbacks({
    onStart: ({ metadata }) => {
      // Track when profile updates begin
      analytics.track("profile_update_started", {
        userId: metadata.prevState?.success
          ? metadata.prevState.data?.user?.id
          : null,
      });
    },
    onSuccess: ({ data }) => {
      revalidatePath("/profile");
      logUserActivity(data.user.id, "profile_updated");
    },
    onError: ({ error }) => {
      if (error.type === "UNHANDLED") {
        logError("Profile update failed", error);
      }
    },
    onSettled: ({ result }) => {
      // Log completion for monitoring and analytics
      analytics.track("profile_update_completed", {
        success: result.success,
      });
    },
  })
  .craft();
```

```typescript
"use client";

import { useActionState } from "react";
import { updateProfile } from "./actions";
import { initial } from "@kellanjs/actioncraft";

export default function ProfileForm() {
  const [state, action] = useActionState(updateProfile, initial(updateProfile));

  return (
    <form action={action}>
      <input
        name="name"
        placeholder="Name"
        defaultValue={state.values?.name}
      />

      <input
        name="email"
        type="email"
        placeholder="Email"
        defaultValue={state.values?.email}
      />

      <textarea
        name="bio"
        placeholder="Bio"
        defaultValue={state.values?.bio}
      />

      {state.success && (
        <div className="success">
          <p>Profile updated successfully!</p>
        </div>
      )}

      {!state.success && state.error.type !== "INITIAL_STATE" && (
        <div className="error">
          {state.error.type === "EMAIL_TAKEN" && (
            <p>That email is already taken. Please use a different one.</p>
          )}
          {state.error.type === "UNAUTHORIZED" && (
            <p>Please log in to update your profile.</p>
          )}
          {state.error.type === "RATE_LIMITED" && (
            <p>Too many requests. Please try again later.</p>
          )}
          {state.error.type === "INPUT_VALIDATION" && (
            <ul>
              {state.error.issues.map((issue, i) => (
                <li key={i}>{issue.message}</li>
              ))}
            </ul>
          )}
        </div>
      )}

      <button type="submit">Update Profile</button>
    </form>
  );
}
```

## Integrations

### Utilities

ActionCraft comes with several utilities intended to make it easier to integrate with libraries like React Query. Let's take a quick look.

#### `ActionCraftError`

A standard Error class that wraps ActionCraft error data while preserving type information:

```typescript
// The error preserves all your action's error data in the `cause` property
if (error instanceof ActionCraftError) {
  console.log(error.message); // "ActionCraft Error: EMAIL_TAKEN - Email already exists"
  console.log(error.cause); // { type: "EMAIL_TAKEN", message: "Email already exists", email: "user@example.com" }
}
```

#### `unwrap(result)`

Extracts the data from a successful result or throws an `ActionCraftError`:

```typescript
const result = await myAction(data);
const userData = unwrap(result); // Throws if result.success === false
```

#### `throwable(action)`

Wraps an action to automatically throw errors as `ActionCraftError` instances instead of returning them:

```typescript
const throwingAction = throwable(myAction);
const userData = await throwingAction(data); // Throws on error
```

#### `isActionCraftError(error, action)`

Type guard that enables full type inference for your action's specific error types:

```typescript
try {
  const data = await throwable(updateUser)(userData);
} catch (error) {
  if (isActionCraftError(error, updateUser)) {
    // error.cause is now typed with updateUser's possible error types
    switch (error.cause.type) {
      case "EMAIL_TAKEN": // Fully typed
      case "UNAUTHORIZED": // Fully typed
      case "INPUT_VALIDATION": // Fully typed
        // Handle each error type
        break;
    }
  }
}
```

**Note:** The `action` argument is required for proper type inference - it tells TypeScript which action's error types to expect.

### React Query

Now let's see how to use these utilities most effectively when working with React Query!

#### Usage with useQuery

Use the `unwrap()` utility for data fetching queries:

```typescript
import { useQuery } from "@tanstack/react-query";
import { fetchUserProfile } from "./actions";
import { unwrap } from "@kellanjs/actioncraft";

function UserProfile({ userId }: { userId: string }) {
  const { data, error, isLoading } = useQuery({
    queryKey: ["user", userId],
    queryFn: async () => {
      const result = await fetchUserProfile({ userId });
      return unwrap(result); // Throws ActionCraftError on failure
    },
  });

  if (isLoading) return <div>Loading...</div>;

  if (error) {
    if (isActionCraftError(error, fetchUserProfile)) {
      // Full type inference for your action's specific error types
      switch (error.cause.type) {
        case "USER_NOT_FOUND":
          return <div>User not found</div>;
        case "UNAUTHORIZED":
          return <div>Please log in</div>;
        default:
          return <div>Error: {error.cause.message}</div>;
      }
    }
    return <div>Unexpected error occurred</div>;
  }

  return (
    <div>
      <h1>{data.user.name}</h1>
      <p>{data.user.email}</p>
    </div>
  );
}
```

If you're like me, and that query function is too verbose for your tastes, you can simplify it:

```typescript
queryFn: () => unwrap(fetchUserProfile({ userId }));
```

`unwrap` is designed to handle both Results and Promises of Results, and since React Query will handle awaiting the resolved Promise, this syntax will work just fine.

#### Usage with useMutation

Use the `throwable()` utility for mutations:

```typescript
import { useMutation, useQueryClient } from "@tanstack/react-query";
import { updateUserProfile } from "./actions";
import { throwable, isActionCraftError } from "@kellanjs/actioncraft";

function EditProfileForm() {
  const queryClient = useQueryClient();

  const mutation = useMutation({
    mutationFn: throwable(updateUserProfile), // Throws ActionCraftError on failure
    onSuccess: (data) => {
      // data is properly typed as your action's success data
      queryClient.invalidateQueries({ queryKey: ["user", data.user.id] });
    },
    onError: (error) => {
      if (isActionCraftError(error, updateUserProfile)) {
        // Handle specific error types with full type safety
        switch (error.cause.type) {
          case "UNAUTHORIZED":
            redirectToLogin();
            break;
          case "INPUT_VALIDATION":
            showValidationErrors(error.cause.issues);
            break;
          case "EMAIL_TAKEN":
            showToast(`Email ${error.cause.email} is already taken`);
            break;
          default:
            showToast(error.cause.message || "Update failed");
        }
      } else {
        showToast("An unexpected error occurred");
      }
    },
  });

  const handleSubmit = (formData: FormData) => {
    mutation.mutate(formData);
  };

  return (
    <form onSubmit={handleSubmit}>
      {/* form fields */}
      <button type="submit" disabled={mutation.isPending}>
        {mutation.isPending ? "Updating..." : "Update Profile"}
      </button>
    </form>
  );
}
```

## Advanced Features

### Bind Arguments

ActionCraft supports binding arguments to actions. Just provide schemas, and you'll get the validated bindArgs values to use in the action.

If validation fails, an error with type "BIND_ARGS_VALIDATION" is returned to the client.

#### Example: Multi-Tenant Action

```typescript
const createPost = create()
  .schemas({
    bindSchemas: [z.string()], // Organization ID
    inputSchema: z.object({
      title: z.string(),
      content: z.string(),
    }),
  })
  .action(async ({ bindArgs, input }) => {
    const [organizationId] = bindArgs;

    const post = await db.post.create({
      data: {
        ...input,
        organizationId,
      },
    });

    return { post };
  })
  .craft();

// Create organization-specific actions
const createPostForOrgA = createPost.bind(null, "org-a-id");
const createPostForOrgB = createPost.bind(null, "org-b-id");

// Each bound action automatically includes the correct org ID
const result = await createPostForOrgA({
  title: "My Post",
  content: "Post content...",
});
```

#### Example: Configuration Binding

```typescript
const sendEmail = create()
  .schemas({
    bindSchemas: [
      z.object({
        apiKey: z.string(),
        fromEmail: z.string(),
      }),
    ],
    inputSchema: z.object({
      to: z.string().email(),
      subject: z.string(),
      body: z.string(),
    }),
  })
  .action(async ({ bindArgs, input }) => {
    const [config] = bindArgs;

    // Use the bound configuration
    const emailService = new EmailService(config.apiKey);
    const result = await emailService.send({
      from: config.fromEmail,
      to: input.to,
      subject: input.subject,
      body: input.body,
    });

    return { messageId: result.id };
  })
  .craft();

// Create environment-specific email actions
const sendProductionEmail = sendEmail.bind(null, {
  apiKey: process.env.PROD_EMAIL_API_KEY,
  fromEmail: "noreply@company.com",
});

const sendDevelopmentEmail = sendEmail.bind(null, {
  apiKey: process.env.DEV_EMAIL_API_KEY,
  fromEmail: "dev@company.com",
});
```

## Type Inference Utilities

ActionCraft provides several type inference utilities to extract types from your actions for use elsewhere in your application:

### Available Utilities

#### `InferInput<Action>`

Infers the input type that the action expects. Returns `unknown` if no input schema is defined.

#### `InferResult<Action>`

Infers the complete result type, including both success and error cases. Respects your action's configuration (api/functional format, useActionState, etc.).

#### `InferData<Action>`

Infers the success data type from your action's return value.

#### `InferErrors<Action>`

Infers all possible error types your action can return, including custom errors, validation errors, and thrown errors.

### Type Extraction Example

```typescript
import { create } from "@kellanjs/actioncraft";
import type {
  InferInput,
  InferResult,
  InferData,
  InferErrors,
} from "@kellanjs/actioncraft";
import { z } from "zod";

const updateUser = create()
  .schemas({
    inputSchema: z.object({
      id: z.string(),
      name: z.string(),
      email: z.string().email(),
    }),
  })
  .errors({
    notFound: (id: string) => ({ type: "NOT_FOUND", id }) as const,
    unauthorized: () => ({ type: "UNAUTHORIZED" }) as const,
  })
  .action(async ({ input, errors }) => {
    // ... implementation
    return { user: input, updatedAt: new Date() };
  })
  .craft();

type ActionInput = InferInput<typeof updateUser>;
// { id: string, name: string, email: string }

type ActionResult = InferResult<typeof updateUser>;
// { success: true, data: { user: UserInput, updatedAt: Date } } |
// { success: false, error: { type: "NOT_FOUND", id: string } | ... }

type ActionData = InferData<typeof updateUser>;
// { user: { id: string, name: string, email: string }, updatedAt: Date }

type ActionErrors = InferErrors<typeof updateUser>;
// { type: "NOT_FOUND", id: string } | { type: "UNAUTHORIZED" } |
// { type: "INPUT_VALIDATION", issues: ... } | { type: "UNHANDLED", message: string }
```

## License

ActionCraft is open source under the terms of the [MIT license](https://github.com/kellanjs/actioncraft/blob/main/LICENSE).