1452: Added chat support with context

drush.services.yml

@@ -19,3 +19,10 @@ services:
       - '@plugin.manager.llm_services'
     tags:
       -  { name: console.command }
+
+  llm.service.char.command:
+    class: Drupal\llm_services\Commands\ModelChatCommand
+    arguments:
+      - '@plugin.manager.llm_services'
+    tags:
+      -  { name: console.command }

src/Client/Ollama.php

@@ -28,7 +28,7 @@ class Ollama {
    * @param string $url
    *   The URL of the Ollama server.
    * @param int $port
-   *   The port that Ollama is listening on.
+   *   The port that Ollama is listening at.
    */
   public function __construct(
     private readonly string $url,
@@ -42,15 +42,13 @@ public function __construct(
    * @return array<string, array<string, string>>
    *   Basic information about the models.
    *
-   * @throws \GuzzleHttp\Exception\GuzzleException
-   * @throws \JsonException
+   * @throws \Drupal\llm_services\Exceptions\CommunicationException
    */
   public function listLocalModels(): array {
     $response = $this->call(method: 'get', uri: '/api/tags');
     $data = $response->getBody()->getContents();
     $data = json_decode($data, TRUE);
 
-    // @todo Change to value objects.
     $models = [];
     foreach ($data['models'] as $item) {
       $models[$item['model']] = [
@@ -135,6 +133,66 @@ public function completion(Payload $payload): \Generator {
     }
   }
 
+  /**
+   * Chat with a model.
+   *
+   * @param \Drupal\llm_services\Model\Payload $payload
+   *   The question to ask the module and the chat history.
+   *
+   * @return \Generator
+   *   The response from the model as it completes it.
+   *
+   * @throws \Drupal\llm_services\Exceptions\CommunicationException
+   * @throws \JsonException
+   */
+  public function chat(Payload $payload): \Generator {
+    $response = $this->call(method: 'post', uri: '/api/chat', options: [
+      'json' => [
+        'model' => $payload->model,
+        'messages' => $this->chatMessagesAsArray($payload),
+        'stream' => TRUE,
+      ],
+      'headers' => [
+        'Content-Type' => 'application/json',
+      ],
+      RequestOptions::CONNECT_TIMEOUT => 10,
+      RequestOptions::TIMEOUT => 300,
+      RequestOptions::STREAM => TRUE,
+    ]);
+
+    $body = $response->getBody();
+    while (!$body->eof()) {
+      $data = $body->read(1024);
+      yield from $this->parse($data);
+    }
+  }
+
+  /**
+   * Take all payload messages and change them into an array.
+   *
+   * This array of messages is used to give the model some chat context to make
+   * the interaction appear more like real char with a person.
+   *
+   * @param \Drupal\llm_services\Model\Payload $payload
+   *   The payload sent to the chat function.
+   *
+   * @return array
+   *   Array of messages to send to Ollama.
+   *
+   * @see https://github.com/ollama/ollama/blob/main/docs/api.md#chat-request-with-history
+   */
+  private function chatMessagesAsArray(Payload $payload): array {
+    $messages = [];
+    foreach ($payload->messages as $message) {
+      $messages[] = [
+        'content' => $message->content,
+        'role' => $message->role->value,
+      ];
+    }
+
+    return $messages;
+  }
+
   /**
    * Parse LLM stream.
    *

src/Client/OllamaChatResponse.php

@@ -0,0 +1,71 @@
+<?php
+
+namespace Drupal\llm_services\Client;
+
+use Drupal\llm_services\Model\ChatResponseInterface;
+use Drupal\llm_services\Model\MessageRoles;
+
+/**
+ * This class represents a completion response in the Ollama provider.
+ */
+readonly class OllamaChatResponse implements ChatResponseInterface {
+
+  /**
+   * Default constructor.
+   *
+   * @param string $model
+   *   Name of the model.
+   * @param string $content
+   *   The content of the message from the model.
+   * @param \Drupal\llm_services\Model\MessageRoles $role
+   *   The role of the message.
+   * @param array $images
+   *   Base64 encoded array of images.
+   * @param bool $done
+   *   The module completion state.
+   */
+  public function __construct(
+    private string $model,
+    private string $content,
+    private MessageRoles $role,
+    private array $images,
+    private bool $done,
+  ) {
+  }
+
+  /**
+   * {@inheritdoc}
+   */
+  public function getModel(): string {
+    return $this->model;
+  }
+
+  /**
+   * {@inheritdoc}
+   */
+  public function getStatus(): bool {
+    return $this->done;
+  }
+
+  /**
+   * {@inheritdoc}
+   */
+  public function getContent(): string {
+    return $this->content;
+  }
+
+  /**
+   * {@inheritdoc}
+   */
+  public function getRole(): MessageRoles {
+    return $this->role;
+  }
+
+  /**
+   * {@inheritdoc}
+   */
+  public function getImages(): array {
+    return $this->images;
+  }
+
+}

src/Commands/ModelChatCommand.php

@@ -0,0 +1,134 @@
+<?php
+
+declare(strict_types=1);
+
+namespace Drupal\llm_services\Commands;
+
+use Drupal\llm_services\Model\Message;
+use Drupal\llm_services\Model\MessageRoles;
+use Drupal\llm_services\Model\Payload;
+use Drupal\llm_services\Plugin\LLModelProviderManager;
+use Symfony\Component\Console\Command\Command;
+use Symfony\Component\Console\Input\InputArgument;
+use Symfony\Component\Console\Input\InputInterface;
+use Symfony\Component\Console\Input\InputOption;
+use Symfony\Component\Console\Output\OutputInterface;
+use Symfony\Component\Console\Question\Question;
+
+/**
+ * Chat with a model through a provider.
+ */
+class ModelChatCommand extends Command {
+
+  /**
+   * Default constructor.
+   *
+   * @param \Drupal\llm_services\Plugin\LLModelProviderManager $providerManager
+   *   The provider manager.
+   */
+  public function __construct(
+    private readonly LLModelProviderManager $providerManager,
+  ) {
+    parent::__construct();
+  }
+
+  /**
+   * {@inheritDoc}
+   */
+  protected function configure(): void {
+    $this
+      ->setName('llm:model:chat')
+      ->setDescription('Chat with model (use ctrl+c to stop chatting)')
+      ->addUsage('llm:model:chat ollama llama3')
+      ->addArgument(
+        name: 'provider',
+        mode: InputArgument::REQUIRED,
+        description: 'Name of the provider (plugin).'
+      )
+      ->addArgument(
+        name: 'name',
+        mode: InputArgument::REQUIRED,
+        description: 'Name of the model to use.'
+      )
+      ->addOption(
+        name: 'system-prompt',
+        mode: InputOption::VALUE_REQUIRED,
+        description: 'System message to instruct the llm have to behave.',
+        default: 'Use the following pieces of context to answer the users question. If you don\'t know the answer, just say that you don\'t know, don\'t try to make up an answer.'
+      )
+      ->addOption(
+        name: 'temperature',
+        mode: InputOption::VALUE_REQUIRED,
+        description: 'The temperature of the model. Increasing the temperature will make the model answer more creatively.',
+        default: '0.8'
+      )
+      ->addOption(
+        name: 'top-k',
+        mode: InputOption::VALUE_REQUIRED,
+        description: 'Reduces the probability of generating nonsense. A higher value (e.g. 100) will give more diverse answers.',
+        default: '40'
+      )
+      ->addOption(
+        name: 'top-p',
+        mode: InputOption::VALUE_REQUIRED,
+        description: 'A higher value (e.g., 0.95) will lead to more diverse text, while a lower value (e.g., 0.5) will generate more focused and conservative text.',
+        default: '0.9'
+      );
+  }
+
+  /**
+   * {@inheritDoc}
+   */
+  protected function execute(InputInterface $input, OutputInterface $output): int {
+    $providerName = $input->getArgument('provider');
+    $name = $input->getArgument('name');
+
+    $systemPrompt = $input->getOption('system-prompt');
+    $temperature = $input->getOption('temperature');
+    $topK = $input->getOption('top-k');
+    $topP = $input->getOption('top-p');
+
+    $provider = $this->providerManager->createInstance($providerName);
+
+    // Build configuration.
+    $payLoad = new Payload();
+    $payLoad->model = $name;
+    $payLoad->options = [
+      'temperature' => $temperature,
+      'top_k' => $topK,
+      'top_p' => $topP,
+    ];
+    $msg = new Message();
+    $msg->role = MessageRoles::System;
+    $msg->content = $systemPrompt;
+    $payLoad->messages[] = $msg;
+
+    $helper = $this->getHelper('question');
+    $question = new Question('Message: ', '');
+
+    // Keep cheating with the user. Not optimal, but okay for now.
+    while (TRUE) {
+      // Query the next question.
+      $output->write("\n");
+      $msg = new Message();
+      $msg->role = MessageRoles::User;
+      $msg->content = $helper->ask($input, $output, $question);
+      $payLoad->messages[] = $msg;
+      $output->write("\n");
+
+      $answer = '';
+      foreach ($provider->chat($payLoad) as $res) {
+        $output->write($res->getContent());
+        $answer .= $res->getContent();
+      }
+      $output->write("\n");
+
+      // Add answer as context to the next question.
+      $msg = new Message();
+      $msg->role = MessageRoles::Assistant;
+      $msg->content = $answer;
+      $payLoad->messages[] = $msg;
+    }
+  }
+
+}

src/Commands/ModelCompletionCommand.php

@@ -14,11 +14,7 @@
 use Symfony\Component\Console\Output\OutputInterface;
 
 /**
- * This is a literal copy of the example Symfony Console command
- * from the documentation.
- *
- * See:
- * http://symfony.com/doc/2.7/components/console/introduction.html#creating-a-basic-command
+ * Make a completion request against a provider model.
  */
 class ModelCompletionCommand extends Command {
 
@@ -41,7 +37,7 @@ protected function configure(): void {
     $this
       ->setName('llm:model:completion')
       ->setDescription('Make a completion request to a model')
-      ->addUsage('llm:model:completion ollama llama2 "Why is the sky blue?')
+      ->addUsage('llm:model:completion ollama llama3 "Why is the sky blue?')
       ->addArgument(
         name: 'provider',
         mode: InputArgument::REQUIRED,

src/Commands/ProviderInstallCommand.php

@@ -4,13 +4,10 @@
 
 namespace Drupal\llm_services\Commands;
 
-use Drupal\llm_services\Model\Message;
-use Drupal\llm_services\Model\Payload;
 use Drupal\llm_services\Plugin\LLModelProviderManager;
 use Symfony\Component\Console\Command\Command;
 use Symfony\Component\Console\Input\InputArgument;
 use Symfony\Component\Console\Input\InputInterface;
-use Symfony\Component\Console\Input\InputOption;
 use Symfony\Component\Console\Output\OutputInterface;
 
 /**
@@ -37,7 +34,7 @@ protected function configure(): void {
     $this
       ->setName('llm:provider:install')
       ->setDescription('Install model in provider')
-      ->addUsage('llm:install:model ollama llama2')
+      ->addUsage('llm:provider:install ollama llama3')
       ->addArgument(
         name: 'provider',
         mode: InputArgument::REQUIRED,

src/Commands/ProviderListCommand.php

@@ -4,13 +4,10 @@
 
 namespace Drupal\llm_services\Commands;
 
-use Drupal\llm_services\Model\Message;
-use Drupal\llm_services\Model\Payload;
 use Drupal\llm_services\Plugin\LLModelProviderManager;
 use Symfony\Component\Console\Command\Command;
 use Symfony\Component\Console\Input\InputArgument;
 use Symfony\Component\Console\Input\InputInterface;
-use Symfony\Component\Console\Input\InputOption;
 use Symfony\Component\Console\Output\OutputInterface;
 
 /**
@@ -37,7 +34,7 @@ protected function configure(): void {
     $this
       ->setName('llm:provider:list')
       ->setDescription('Install model in provider')
-      ->addUsage('llm:install:model ollama llama2')
+      ->addUsage('llm:provider:list ollama')
       ->addArgument(
         name: 'provider',
         mode: InputArgument::REQUIRED,