Skip to content

ai_connector.md

Latest commit

 

History

History
96 lines (67 loc) · 2.29 KB

ai_connector.md

File metadata and controls

96 lines (67 loc) · 2.29 KB
Raw

AI Connector

Engine\Atomic\Tools\AIConnector is a small wrapper around chat-completions and embeddings APIs.

Supported providers:

  • openai
  • groq
  • openrouter
  • globus
$ai = ai_connector('your-api-key');

$text = $ai->completion('Explain queues in one paragraph.');

$result = $ai->chat_completion([
    ['role' => 'system', 'content' => 'You are concise.'],
    ['role' => 'user', 'content' => 'List three cache strategies.'],
]);

echo $result->choices[0]->message->content ?? '';

Provider and model

$ai = ai_connector(null, 'openai'); // reads ai.openai.api_key from config

$ai->set_provider('openrouter')
   ->set_model('google/gemini-2.5-flash');

If no model is set manually, the connector picks a provider-specific default:

  • OpenAI: gpt-5-nano
  • Groq: llama-3.1-8b-instant
  • OpenRouter: google/gemini-2.0-flash-001
  • Globus: globus-general

Available models

$models = $ai->get_available_models();
print_r($models);

get_available_models() returns a curated static list from the connector class. It does not fetch models remotely.

Chat options

chat_completion() accepts an optional second argument:

$result = $ai->chat_completion($messages, [
    'model' => 'gpt-5-mini',
    'temperature' => 0.2,
    'max_tokens' => 800,
    'timeout' => 60,
]);

Provider-specific options:

  • OpenRouter: referer, user_agent, lang

Embeddings

$embeddings = $ai->create_embeddings('Atomic framework docs');
$vector = $embeddings->data[0]->embedding ?? [];

$batch = $ai->create_embeddings([
    'First text',
    'Second text',
]);

create_embeddings() sends requests to the /embeddings endpoint and defaults to text-embedding-3-large.

Config

The connector reads API keys from the Atomic hive when no key is passed explicitly:

$atomic->set('ai.openai.api_key', 'sk-...');
$atomic->set('ai.groq.api_key', 'gsk-...');
$atomic->set('ai.openrouter.api_key', 'or-...');
$atomic->set('ai.globus.api_key', '...');

Notes

  • ai_connector() returns a shared singleton instance.
  • Switching provider on that instance also resets the default model for the selected provider.
  • The current implementation exposes chat completions, plain prompt completion, and embeddings only. It does not implement streaming helpers.