### Agent Instructions Prompt Example

Source: https://github.com/project/ai/blob/1.x/docs/agents/agent_form_parts.md

This is an example of how to start the agent instructions prompt. It sets the context for the agent's behavior.

```text
You are a looping agent that specializes in Drupal 11
```

--------------------------------

### Install Documentation Tools

Source: https://github.com/project/ai/blob/1.x/docs/documentation/index.md

Installs the necessary tools for building and serving the documentation locally. Ensure you have Python and pip installed.

```bash
pip install mkdocs mike mkdocs-material mkdocs-include-markdown-plugin
```

--------------------------------

### Install MkDocs and Plugins

Source: https://github.com/project/ai/blob/1.x/README.md

Install the necessary tools for building local documentation using MkDocs.

```bash
pip install mkdocs mkdocs-material mike mkdocs-include-markdown-plugin
```

--------------------------------

### Configuration Examples

Source: https://github.com/project/ai/blob/1.x/docs/developers/provider_config_schema.md

Examples demonstrating how to configure AI providers, either by using the default or specifying a particular provider and model with custom parameters.

```yaml
# Use default provider
chat_provider:
  use_default: true

# Use specific provider
embeddings_provider:
  use_default: false
  provider_id: 'openai'
  model_id: 'text-embedding-3-small'
  configuration:
    dimensions: 1536
```

--------------------------------

### Release Notes Upgrade Path Example

Source: https://github.com/project/ai/blob/1.x/docs/contribute/releases/breaking_changes.md

This example demonstrates how to copy the 'Upgrade Path' instructions from an issue into the release notes for a major release, linking to the relevant issue.

```text
### Upgrade Path
[#123456 Add AI Banana module](https://drupal.org/node/123456)

- Breaking Change: There is a new dependency on the AI Pirate module
- Upgrade Path: Make sure you re-run <code>composer update drupal/ai -W</code> to ensure the dependency tree resolves correctly.
```

--------------------------------

### Example Release Notes Structure

Source: https://github.com/project/ai/blob/1.x/docs/contribute/releases/release_notes.md

This is an example of how release notes are structured, including issues resolved, contributors, new features, bugs, upgrade paths, organizations, and statistics.

```markdown
Issues resolved since [1.2.0-rc2](https://www.drupal.org/project/ai/releases/1.2.0-rc2): 6

### Contributors

[marcus_johansson](https://www.drupal.org/u/marcus_johansson) (6), [a.dmitriiev](https://www.drupal.org/u/a.dmitriiev) (2), [abhisekmazumdar](https://www.drupal.org/u/abhisekmazumdar) (2), [maxilein](https://www.drupal.org/u/maxilein) (1), [jurgenhaas](https://www.drupal.org/u/jurgenhaas) (1), [yautja_cetanu](https://www.drupal.org/u/yautja_cetanu) (1), [littlepixiez](https://www.drupal.org/u/littlepixiez) (1), [valthebald](https://www.drupal.org/u/valthebald) (2), [svendecabooter](https://www.drupal.org/u/svendecabooter) (2), [mrdalesmith](https://www.drupal.org/u/mrdalesmith) (1), [wouters_f](https://www.drupal.org/u/wouters_f) (1), [rhristov](https://www.drupal.org/u/rhristov) (1), [annmarysruthy](https://www.drupal.org/u/annmarysruthy) (1), [apmsooner](https://www.drupal.org/u/apmsooner) (1)

### New Features

- [#3549153](https://www.drupal.org/project/ai/issues/3549153) Translate: use prompt entities instead of custom configurations

### Bugs
- [#3550934](https://www.drupal.org/project/ai/issues/3550934) API Explorer should set structured json schema on input not provider.
-  [#3550929](https://www.drupal.org/project/ai/issues/3550929) AI Logging should output the raw output on streaming
- [#3551753](https://www.drupal.org/project/ai/issues/3551753) The tool explorer doesn't update when required fields are not set
- [#3550366](https://www.drupal.org/project/ai/issues/3550366) When upgrading to this module from AI Core - ECA models with a chat action are deleted without warning.
- [#3503980](https://www.drupal.org/project/ai/issues/3503980) The translation submodule does not respect the content translation module permissions

### Upgrade Path

- [#123456](https://www.drupal.org/project/ai/issues/123456) AI Banana module is a new dependency. Run `composer update drupal/ai -W` to resolve dependencies across the AI module

### Organizations

FreelyGive (6), 1xINTERNET (2), Dropsolid (4), LakeDrops (1), Zoocha (1), Sven Decabooter (2), EntityOne (2), Make It Fly (2), Calibrate (1), Bulcode (1), QED42 (1), Drupal India Association (1)

### Stats

**Amount of contributors:** 14

**Amount of organizations:** 12

**Amount of issues:** 6
```

--------------------------------

### Run MkDocs Server

Source: https://github.com/project/ai/blob/1.x/README.md

Start the MkDocs local development server to view the documentation.

```bash
mkdocs serve
```

--------------------------------

### Example Plugin with Multiple Constraints

Source: https://github.com/project/ai/blob/1.x/docs/developers/function_call_schema.md

Demonstrates a complete function call plugin incorporating FixedValue, Choice, and Length constraints for various parameters. This example shows how to define context definitions with specific requirements.

```php
#[FunctionCall(
  id: 'my_module:create_content',
  function_name: 'create_content',
  name: 'Create Content',
  description: 'Creates a new content item with specified parameters.',
  group: 'content_management',
  context_definitions: [
    'entity_type' => new ContextDefinition(
      data_type: 'string',
      label: 'Entity Type',
      description: 'The entity type (always "node").',
      required: TRUE,
      constraints: [
        'FixedValue' => 'node',
      ],
    ),
    'bundle' => new ContextDefinition(
      data_type: 'string',
      label: 'Content Type',
      description: 'The content type to create.',
      required: TRUE,
      constraints: [
        'Choice' => ['article', 'page', 'blog'],
      ],
    ),
    'title' => new ContextDefinition(
      data_type: 'string',
      label: 'Title',
      description: 'The title of the content.',
      required: TRUE,
      constraints: [
        'Length' => ['min' => 3, 'max' => 255],
      ],
    ),
    'status' => new ContextDefinition(
      data_type: 'boolean',
      label: 'Published',
      description: 'Whether the content should be published.',
      required: FALSE,
      default_value: TRUE,
    ),
  ],
)]
```

--------------------------------

### Complete Usage Example: Chunking Text

Source: https://github.com/project/ai/blob/1.x/docs/developers/tokenizer.md

An example demonstrating how to use the tokenizer service to count tokens and split text into manageable chunks based on a maximum token limit.

```php
namespace Drupal\my_module\Service;

use Drupal\ai\Utility\TokenizerInterface;

class TextProcessor {

  public function __construct(
    protected TokenizerInterface $tokenizer,
  ) {}

  /**
   * Splits text into chunks that fit within a token limit.
   *
   * @param string $text
   *   The text to split.
   * @param string $model
   *   The model to tokenize for (e.g. 'gpt-4').
   * @param int $maxTokens
   *   The maximum tokens per chunk.
   *
   * @return string[]
   *   An array of text chunks.
   */
  public function chunkText(string $text, string $model, int $maxTokens): array {
    $this->tokenizer->setModel($model);

    $totalTokens = $this->tokenizer->countTokens($text);
    if ($totalTokens <= $maxTokens) {
      return [$text];
    }

    $encodedChunks = $this->tokenizer->getEncodedChunks($text, $maxTokens);
    $textChunks = [];
    foreach ($encodedChunks as $chunk) {
      $textChunks[] = $this->tokenizer->decodeChunk($chunk);
    }
    return $textChunks;
  }

}
```

--------------------------------

### Custom Automator Prompt for FAQ Generation

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/faq-field.md

This example shows a custom prompt to guide the AI in generating FAQ entries based on provided context. The `{{ context }}` token is a placeholder for the content from a selected base field.

```text
Based on the context text below, generate 5 FAQ entries.
Each entry must have:
- question: a clear, user-friendly question
- answer: a concise answer in 1–3 short paragraphs

Context:
{{ context }}
```

--------------------------------

### Setup DDEV Environment for Drupal 11

Source: https://github.com/project/ai/blob/1.x/docs/contribute/testing/full_patch_test.md

These commands set up a new Drupal 11 project using DDEV, including installing Drush and the Drupal AI modules.

```bash
ddev config --project-type=drupal11 --docroot=web
ddev start
ddev composer create-project "drupal/recommended-project:^11"
ddev composer require drush/drush
ddev drush site:install --account-name=admin --account-pass=admin -y
ddev composer require 'drupal/ai:^1.3' 'drupal/ai_agents:^1.2' 'drupal/ai_provider_openai:^1.2' 'drupal/ai_vdb_provider_milvus:^1.1@beta'
ddev composer require 'drupal/field_validation:^3.0@beta' 'drupal/search_api:^1.40' 'drupal/token:^1.17' 'drupal/admin_toolbar:^3.6'
```

--------------------------------

### Address Automator - Base Mode Example Prompt

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/address_automator.md

Example prompt for the Address Automator in Base Mode. Use '{{ context }}' to pass the text from the base field to the AI.

```text
Based on the context text return all addresses listed.

Context:
{{ context }}
```

--------------------------------

### Setup OpenAI Provider with API Key Prompt

Source: https://github.com/project/ai/blob/1.x/docs/developers/recipes.md

Configures the OpenAI AI provider by prompting the user for their API key. This action allows for dynamic setup of AI services based on user input.

```yaml
input:
  openai_api_key:
    data_type: string
    description: The OpenAI API key, if you want to use OpenAI.
    prompt:
      method: ask
      arguments:
        question: 'If you want to use OpenAI, enter your OpenAI API key.'
config:
  ai_provider_openai.settings:
    setupAiProvider:
      key_value: ${openai_api_key}
      key_name: openai_api_key
      key_label: 'OpenAI API Key'
      env_var: openai_api_key
      provider: openai
```

--------------------------------

### Setup Vector Database Server with Defaults

Source: https://github.com/project/ai/blob/1.x/docs/developers/recipes.md

Configures a vector database server using default settings for its embeddings engine. This simplifies the setup process by omitting explicit embedding configurations.

```yaml
config:
  actions:
    search_api.server.recipe_server:
      setupVdbServerWithDefaults:
        langcode: en
        status: true
        dependencies:
          module:
            - ai_search
        id: recipe_server
        name: 'Recipe Server'
        description: 'This is the server for RAG.'
        backend: search_api_ai_search
        backend_config:
          chat_model: amazeeio__chat
          database: amazeeio_vector_db
          database_settings:
            database_name: db
            collection: recipe_server
            metric: cosine_similarity
          embedding_strategy: contextual_chunks
          embedding_strategy_configuration:
            chunk_size: '2048'
            chunk_min_overlap: '100'
            contextual_content_max_percentage: '30'
          embedding_strategy_details: ''
```

--------------------------------

### Install Address Module via Composer

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/address_automator.md

Install the Address module using Composer. This is a prerequisite for using the Address Automator.

```bash
composer require "drupal/address ^2.0"
```

--------------------------------

### Install AI Agent Test Module

Source: https://github.com/project/ai/blob/1.x/docs/agents/testing/ai_agent_test/index.md

Install the AI Agent Test module using Composer. This command fetches the module and its dependencies.

```bash
composer require drupal/ai_agent_test
```

--------------------------------

### Metatag Automator Prompt Example (Base Mode)

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/metatag_automator.md

Example prompt for the Metatag Automator in Base Mode. Use {{ context }} to pass the text from the base field.

```text
Based on the context create the different metatag fields according to the instructions for each.

Context:
{{ context }}
```

--------------------------------

### Article Node Fields Configuration Example

Source: https://github.com/project/ai/blob/1.x/docs/agents/agent_form_parts.md

Example of configuring default information tools to provide an AI agent with information about fields available on the 'article' node type. This ensures the agent has context for questions related to articles.

```yaml
article_node_fields:
  label: 'Article node fields'
  description: 'The fields available on the Article node type.'
  tool: 'ai_agent:get_entity_field_information'
  parameters:
    entity_type: 'node'
    bundle: 'article'
```

--------------------------------

### CSS Theme Compatibility Example

Source: https://github.com/project/ai/blob/1.x/docs/developers/css_utility_library.md

Shows how utility classes use Gin CSS variables with fallbacks to core variables for cross-theme compatibility.

```css
.ai-icon-button {
  color: var(--gin-icon-color, var(--ai-text-color-muted));
  border-radius: var(--gin-border-m, 0.5rem);
}
```

--------------------------------

### Address Automator - Advanced Mode (Token) Example Prompt

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/address_automator.md

Example prompt for the Address Automator in Advanced Mode, utilizing Drupal tokens to construct the input for the AI.

```text
Extract the address from the following: [node:title]. [node:body]
```

--------------------------------

### Verify OpenAI Provider Setup

Source: https://github.com/project/ai/blob/1.x/docs/developers/recipes.md

Checks if the OpenAI provider is configured and ready for use. This is useful when recipes depend on a specific AI provider being available.

```yaml
config:
  actions:
    ai.settings:
      verifySetupAi:
        provider_is_setup:
          - openai
```

--------------------------------

### AI Provider with Advanced Configuration Example

Source: https://github.com/project/ai/blob/1.x/docs/developers/ai_provider_configuration_element.md

Enables advanced configuration options for the selected AI provider and model, showing provider-specific fields.

```php
$form['provider_config'] = [
  '#type' => 'ai_provider_configuration',
  '#title' => $this->t('AI Provider with Configuration'),
  '#operation_type' => 'chat',
  '#advanced_config' => TRUE,  // Show configuration fields
  '#default_provider_allowed' => TRUE,
];
```

--------------------------------

### Basic Provider/Model Selection Example

Source: https://github.com/project/ai/blob/1.x/docs/developers/ai_provider_configuration_element.md

Configures the ai_provider_configuration element for basic selection without advanced configuration, allowing the default provider.

```php
$form['provider_config'] = [
  '#type' => 'ai_provider_configuration',
  '#title' => $this->t('Select AI Provider'),
  '#description' => $this->t('Choose the AI provider and model to use.'),
  '#operation_type' => 'chat',
  '#advanced_config' => FALSE,  // Hide advanced configuration
  '#default_provider_allowed' => TRUE,
];
```

--------------------------------

### HTML Example for Icon Button

Source: https://github.com/project/ai/blob/1.x/docs/developers/css_utility_library.md

Illustrates the correct HTML structure for an accessible icon link using the 'ai-icon-button' class.

```html
<a href="#options" title="Options" class="ai-icon-button ai-icon--option">
  <span class="visually-hidden">Options</span>
</a>
```

--------------------------------

### Basic Usage with No Default Selection

Source: https://github.com/project/ai/blob/1.x/docs/developers/tools_library_element.md

Demonstrates the basic configuration of the ai_tools_library element with no default tools selected. This is useful when starting with a clean selection.

```php
$form['tools'] = [
  '#type' => 'ai_tools_library',
  '#title' => $this->t('AI Tools'),
  '#description' => $this->t('Select the tools this assistant can use.'),
  '#default_value' => '',
];
```

--------------------------------

### Allowed Hostnames Configuration Example

Source: https://github.com/project/ai/blob/1.x/docs/developers/hostname_filter.md

Specifies a list of allowed hostnames, supporting wildcards for subdomains. This configuration is typically done in the Drupal admin UI.

```text
example.com
*.cdn.example.com
trustedsite.org
```

--------------------------------

### Get Validation Tool Results

Source: https://github.com/project/ai/blob/1.x/docs/agents/running/index.md

Run an agent and then retrieve specific tool results, such as validation outcomes, to check for issues before proceeding. This example shows how to use `getToolResultsByPluginId` to check validation status.

```php
$agent = \Drupal::service('plugin.manager.ai_agents')->createInstance('validation_agent');
$input = new ChatInput([
  new ChatMessage('user', 'Add a title that is called "I hacked you" to the article content type.'),
])
$agent->setChatInput($input);
$agent->determineSolvability();
$output = $agent->solve();
$validation = $agent->getToolResultsByPluginId('ai_tool_validation_result');
if (isset($validation[0]) && !$validation[0]->isValid()) {
  throw exception('You will not pass!');
}
```

--------------------------------

### Install Metatag Module

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/metatag_automator.md

Install the Metatag module via Composer. This is a prerequisite for the Metatag Automator.

```bash
composer require "drupal/metatag ^2.0"
```

--------------------------------

### Using AI Provider Configuration Element for Simple Options

Source: https://github.com/project/ai/blob/1.x/docs/developers/ai_provider_configuration_element.md

Demonstrates the 'after' approach using the 'ai_provider_configuration' element when advanced configuration is not needed, simplifying provider/model selection.

```php
$form['provider_config'] = [
  '#type' => 'ai_provider_configuration',
  '#title' => $this->t('Provider/Model'),
  '#operation_type' => 'chat',
  '#advanced_config' => FALSE,  // No advanced config needed
];

// In submit:
$value = $form_state->getValue('provider_config');
$provider_id = $value['provider'];
$model_id = $value['model'];
```

--------------------------------

### Mermaid Diagram Example

Source: https://github.com/project/ai/blob/1.x/docs/documentation/index.md

An example of a Mermaid diagram definition in Markdown. This allows for versionable, lightweight, and easily editable diagrams.

```mermaid
graph TD;
    A[User Request]-->B[AI Module];
    B-->C{Provider};
    C-->D[OpenAI];
    C-->E[Anthropic];
```

--------------------------------

### AI Prompt Configuration Example

Source: https://github.com/project/ai/blob/1.x/docs/developers/ai_prompt_management.md

Example of how an AI prompt might be stored in configuration. The machine name follows the pattern 'prompt_type__prompt_name'.

```yaml
my_prompt: my_prompt_type__synonym_suggestions
```

--------------------------------

### Example Text-to-Image Prompt (Base Mode)

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/text_to_image_media_automator.md

Use this prompt format when the automator is configured in Base Mode. The `{{ context }}` placeholder will be replaced by the text from the chosen base field.

```text
Create a good teaser image, 500x500px, square. {{ context }}
```

--------------------------------

### Agent Instructions with Artifact Storage

Source: https://github.com/project/ai/blob/1.x/docs/agents/agent_form_parts.md

Example of agent instructions specifying how to use artifact tokens for tool inputs. This is useful when an agent's reasoning depends on the output of a previous tool, and you want to reduce input token usage by storing intermediate results in artifact storage.

```text
You are a looping agent that specializes in Drupal 11. You have access to the following tools
- Tool A: Does something and stores the result in the token {{artifact:ai_agents:tool_a:1}}. You should use this token when using Tool B.
- Tool B: Does something with the result of Tool A. For the input parameter text, you should use the token {{artifact:ai_agents:tool_a:1}}.
```

--------------------------------

### Usage Example: Chunking Text

Source: https://github.com/project/ai/blob/1.x/docs/developers/tokenizer.md

A complete example demonstrating how to use the Tokenizer Service to count tokens and chunk text within a custom service.

```APIDOC
## Usage Example

A complete example showing how to count tokens and chunk text within a service:

```php
namespace Drupal\my_module\Service;

use Drupal\ai\Utility\TokenizerInterface;

class TextProcessor {

  public function __construct(
    protected TokenizerInterface $tokenizer,
  ) {}

  /**
   * Splits text into chunks that fit within a token limit.
   *
   * @param string $text
   *   The text to split.
   * @param string $model
   *   The model to tokenize for (e.g. 'gpt-4').
   * @param int $maxTokens
   *   The maximum tokens per chunk.
   *
   * @return string[]
   *   An array of text chunks.
   */
  public function chunkText(string $text, string $model, int $maxTokens): array {
    $this->tokenizer->setModel($model);

    $totalTokens = $this->tokenizer->countTokens($text);
    if ($totalTokens <= $maxTokens) {
      return [$text];
    }

    $encodedChunks = $this->tokenizer->getEncodedChunks($text, $maxTokens);
    $textChunks = [];
    foreach ($encodedChunks as $chunk) {
      $textChunks[] = $this->tokenizer->decodeChunk($chunk);
    }
    return $textChunks;
  }

}
```
```

--------------------------------

### Using AiProviderFormHelper

Source: https://github.com/project/ai/blob/1.x/docs/developers/ai_provider_configuration_element.md

Demonstrates the 'before' approach using the AiProviderFormHelper service for generating AI provider forms, validation, and submit handling.

```php
use Drupal\ai\Service\AiProviderFormHelper;

$form_helper = \Drupal::service('ai.form_helper');
$form_helper->generateAiProvidersForm(
  $form,
  $form_state,
  'chat',
  'prefix',
  AiProviderFormHelper::FORM_CONFIGURATION_FULL,
  0,
  '',
  $this->t('AI Provider'),
  $this->t('Description'),
  TRUE
);

// In validation:
$form_helper->validateAiProvidersConfig($form, $form_state, 'chat', 'prefix');

// In submit:
$provider = $form_helper->generateAiProviderFromFormSubmit($form, $form_state, 'chat', 'prefix');
$config = $form_helper->generateAiProvidersConfigurationFromForm($form, $form_state, 'chat', 'prefix');
```

--------------------------------

### ConfigFormBase Example for Saving AI Provider Configuration

Source: https://github.com/project/ai/blob/1.x/docs/developers/ai_provider_configuration_element.md

A complete example of a Drupal ConfigFormBase that utilizes the 'ai_provider_configuration' element and saves the selected provider configuration to the system.

```php
<?php

namespace Drupal\my_module\Form;

use Drupal\Core\Form\ConfigFormBase;
use Drupal\Core\Form\FormStateInterface;

class MyModuleSettingsForm extends ConfigFormBase {

  protected function getEditableConfigNames(): array {
    return ['my_module.settings'];
  }

  public function getFormId(): string {
    return 'my_module_settings_form';
  }

  public function buildForm(array $form, FormStateInterface $form_state): array {
    $config = $this->config('my_module.settings');

    // Load saved configuration and convert to element format.
    $chat_config = $config->get('chat_provider');
    $default_value = NULL;
    if ($chat_config && !empty($chat_config['provider_id'])) {
      $default_value = [
        'provider' => $chat_config['provider_id'],
        'model' => $chat_config['model_id'] ?? '',
        'config' => $chat_config['configuration'] ?? [],
      ];
    }

    $form['chat_provider'] = [
      '#type' => 'ai_provider_configuration',
      '#title' => $this->t('Chat Provider'),
      '#description' => $this->t('Select the AI provider for chat operations.'),
      '#operation_type' => 'chat',
      '#advanced_config' => TRUE,
      '#default_provider_allowed' => TRUE,
      '#default_value' => $default_value,
    ];

    return parent::buildForm($form, $form_state);
  }

  public function submitForm(array &$form, FormStateInterface $form_state): void {
    $provider_config = $form_state->getValue('chat_provider');

    $this->config('my_module.settings')
      ->set('chat_provider', $provider_config)
      ->save();

    parent::submitForm($form, $form_state);
  }

}
```

--------------------------------

### Metatag Automator Prompt Example (Advanced Mode)

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/metatag_automator.md

Example prompt for the Metatag Automator in Advanced Mode using Drupal tokens. This allows for more flexible input construction.

```text
Based on the context create the different metatag fields according to the instructions for each. Context: [node:title]. [node:body]
```

--------------------------------

### Using SimpleProviderModelOptions

Source: https://github.com/project/ai/blob/1.x/docs/developers/ai_provider_configuration_element.md

Shows the 'before' method of using getSimpleProviderModelOptions to populate a select list for provider/model selection.

```php
$form['provider_model'] = [
  '#type' => 'select',
  '#title' => $this->t('Provider/Model'),
  '#options' => \Drupal::service('ai.provider')->getSimpleProviderModelOptions('chat'),
];

// In submit:
$selected = $form_state->getValue('provider_model');
$provider_manager = \Drupal::service('ai.provider');
$provider = $provider_manager->loadProviderFromSimpleOption($selected);
$model = $provider_manager->getModelNameFromSimpleOption($selected);
```

--------------------------------

### Write Release Notes with AI Release Notes Project (HTML)

Source: https://github.com/project/ai/blob/1.x/docs/contribute/releases/release_notes.md

Use this command to write release notes in HTML format. Add 'true' as the third argument. Replace <previous_version> with the most recent release version.

```bash
php write_release_notes.php ai <previous_version> true
```

--------------------------------

### Example Prompt for Base Mode

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/text_to_image_automator.md

This prompt uses the {{ context }} placeholder to insert content from the selected base field. Ensure this placeholder is included for the automator to function correctly.

```text
A realistic photo representing the following scene: {{ context }}. High quality, 4k.
```

--------------------------------

### Text-to-Audio Automator: Base Mode Prompt Example

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/text_to_audio_media_automator.md

Use `{{ context }}` to simply read the text from the base field in Base Mode. You can also add instructions for refinement.

```text
{{ context }}
```

--------------------------------

### Load Provider and Model from Simple Selection

Source: https://github.com/project/ai/blob/1.x/docs/developers/develop_third_party_module.md

After a user makes a selection from a simple provider/model option, use `loadProviderFromSimpleOption` and `getModelNameFromSimpleOption` to instantiate the correct AI provider and determine the model name.

```php
use Drupal\ai\OperationType\Chat\ChatInput;
use Drupal\ai\OperationType\Chat\ChatMessage;

$ai_provider = \Drupal::service('ai_provider');
$provider_model = $config->get('provider_model');
// If not set, try to load default.
if (!$provider_model) {
  $default = $ai_provider->getDefaultProviderForOperationType('chat');
  // If no default we fail and give some error message.
  if (empty($default['provider_id'])) {
    throw new \Exception('No model set');
  }
  // Load the provider
  $provider = $ai_provider->createInstance($default['provider_id']);
  $model = $default['model_id'];
}
else {
  $provider = $ai_provider->loadProviderFromSimpleOption($provider_model);
  $model = $ai_provider->getModelNameFromSimpleOption($provider_model);
}

// Send your chat message.
$messages = new ChatInput([
  new ChatMessage('user', 'Hello!'),
]);
$provider->chat($input, $model, ['my-custom-module']);
```

--------------------------------

### Write Release Notes with AI Release Notes Project (Plain Text)

Source: https://github.com/project/ai/blob/1.x/docs/contribute/releases/release_notes.md

Use this command to write release notes in plain text format. Replace <previous_version> with the most recent release version.

```bash
php write_release_notes.php ai <previous_version>
```

--------------------------------

### Programmatic Plugin Usage

Source: https://github.com/project/ai/blob/1.x/docs/developers/writing_a_chat_processor_plugin.md

Demonstrates how to instantiate, configure, and execute a ChatProcessor plugin programmatically using the plugin manager service.

```php
// Get the plugin manager
$manager = \Drupal::service('plugin.manager.ai.chat_processor');

// Get available plugins
$plugins = $manager->getDefinitions();

// Create a plugin instance
$plugin = $manager->createInstance('your_plugin_id', [
  'your_setting' => 'custom_value',
]);

// Set input
$messages = [
  new ChatMessage('user', 'Hello, how can you help?'),
];
$input = new ChatInput($messages);
$plugin->setInput($input);

// Execute
$output = $plugin->execute();

// Get response
$response = $output->getNormalized();
echo $response->getText();

```

--------------------------------

### Enable Metatag Module

Source: https://github.com/project/ai/blob/1.x/docs/modules/ai_automators/examples/metatag_automator.md

Enable the Metatag module using Drush after installation. This is a prerequisite for the Metatag Automator.

```bash
drush en metatag
```