| Server IP : 121.121.20.254 / Your IP : 216.73.217.64 Web Server : Microsoft-IIS/10.0 System : Windows NT WEB-SERVER 10.0 build 20348 (Windows Server 2022) AMD64 User : IUSR ( 0) PHP Version : 8.3.28 Disable Function : NONE MySQL : ON | cURL : ON | WGET : OFF | Perl : OFF | Python : OFF | Sudo : OFF | Pkexec : OFF Directory : C:/inetpub/wwwroot/AIWEBSTATION/wp-content/plugins/ai-engine/classes/engines/ |
Upload File : |
<?php
class Meow_MWAI_Engines_OVH extends Meow_MWAI_Engines_ChatML {
private $endpoint = 'https://oai.endpoints.kepler.ai.cloud.ovh.net';
public $envType = 'ovh';
public function __construct( $core, $env ) {
parent::__construct( $core, $env );
}
/**
* Sets up the environment for our custom engine.
*/
protected function set_environment() {
$env = $this->env;
if ( !isset( $env['apikey'] ) || empty( $env['apikey'] ) ) {
throw new Exception( 'AI Engine: OVH API key is not set. Please configure the OVH AI Endpoints access token in your settings.' );
}
$this->apiKey = $env['apikey'];
if ( isset( $env['endpoint'] ) && !empty( $env['endpoint'] ) ) {
$this->endpoint = $env['endpoint'];
}
}
protected function build_url( $query, $endpoint = null ) {
$endpoint = apply_filters( 'mwai_ovh_endpoint', trailingslashit( $this->endpoint ) . 'v1', $this->env );
return $endpoint . '/chat/completions';
}
protected function get_service_name() {
return 'OVH';
}
protected function build_messages( $query ) {
// Handle vision models like llava-next-mistral-7b
if ( strpos( $query->model, 'llava' ) !== false ) {
$query->image_remote_upload = 'data';
}
$messages = parent::build_messages( $query );
return $messages;
}
protected function stream_data_handler( $json ) {
// Handle usage-only chunks (final chunk with empty choices array).
// OVH sends usage data in a final chunk with choices: []
if ( isset( $json['usage'] ) && empty( $json['choices'] ) ) {
$usage = $json['usage'];
if ( isset( $usage['prompt_tokens'], $usage['completion_tokens'] ) ) {
$this->streamInTokens = (int) $usage['prompt_tokens'];
$this->streamOutTokens = (int) $usage['completion_tokens'];
}
return null; // No content in this chunk
}
// Let parent handle all other chunks
return parent::stream_data_handler( $json );
}
public function get_models() {
return $this->retrieve_models();
}
protected function build_headers( $query ) {
$headers = [
'Content-Type' => 'application/json',
'Authorization' => 'Bearer ' . $this->apiKey,
];
return $headers;
}
protected function build_body( $query, $streamCallback = null, $extra = null ) {
$body = parent::build_body( $query, $streamCallback, $extra );
// Handle max_tokens parameter (OVH uses max_tokens, not max_completion_tokens)
if ( isset( $body['max_completion_tokens'] ) ) {
$body['max_tokens'] = $body['max_completion_tokens'];
unset( $body['max_completion_tokens'] );
}
// OVH does not error when max_tokens cannot fit: it returns 200 with an
// instantly-terminated stream (a lone [DONE]), which looked like the AI
// answering nothing. Its models also share ONE window between prompt and
// completion (max output == context), so any max_tokens at or above the
// model's cap can never fit once the prompt is counted. In that case drop
// the parameter and let the endpoint default to the remaining context.
if ( !empty( $body['max_tokens'] ) ) {
$modelInfo = $this->retrieve_model_info( $query->model );
if ( !empty( $modelInfo['maxCompletionTokens'] ) &&
(int) $body['max_tokens'] >= (int) $modelInfo['maxCompletionTokens'] ) {
unset( $body['max_tokens'] );
}
}
// OVH supports stream_options for accurate token usage in streaming
if ( !empty( $streamCallback ) && !isset( $body['stream_options'] ) ) {
$body['stream_options'] = [
'include_usage' => true,
];
}
// Remove parallel_tool_calls - OVH might not support it
if ( isset( $body['parallel_tool_calls'] ) ) {
unset( $body['parallel_tool_calls'] );
}
// Remove other potentially unsupported parameters
$unsupported_params = [ 'response_format', 'seed', 'logit_bias', 'logprobs', 'top_logprobs' ];
foreach ( $unsupported_params as $param ) {
if ( isset( $body[$param] ) ) {
unset( $body[$param] );
}
}
return $body;
}
public function handle_tokens_usage(
$reply,
$query,
$returned_model,
$returned_in_tokens,
$returned_out_tokens,
$returned_price = null
) {
// Clean up the data
$returned_in_tokens = !is_null( $returned_in_tokens ) ?
$returned_in_tokens : $reply->get_in_tokens( $query );
$returned_out_tokens = !is_null( $returned_out_tokens ) ?
$returned_out_tokens : $reply->get_out_tokens();
// Calculate price based on our model definitions
$models = $this->get_ovh_models();
$model_price = null;
foreach ( $models as $model ) {
if ( $model['model'] === $returned_model ) {
$model_price = $model['price'];
break;
}
}
if ( $model_price ) {
$returned_price = ( $returned_in_tokens * $model_price['in'] +
$returned_out_tokens * $model_price['out'] ) / 1000000;
}
else {
$returned_price = 0;
}
// Record the usage
$usage = $this->core->record_tokens_usage(
$returned_model,
$returned_in_tokens,
$returned_out_tokens,
$returned_price
);
// Set the usage in the reply
$reply->set_usage( $usage );
// Set accuracy based on data availability
if ( !is_null( $returned_in_tokens ) && !is_null( $returned_out_tokens ) ) {
// Tokens from API (via stream_options), price calculated locally
$reply->set_usage_accuracy( 'tokens' );
}
else {
// Everything estimated
$reply->set_usage_accuracy( 'estimated' );
}
}
public function get_price( Meow_MWAI_Query_Base $query, Meow_MWAI_Reply $reply ) {
$models = $this->get_ovh_models();
foreach ( $models as $model ) {
if ( $model['model'] === $query->model ) {
$in_tokens = $reply->get_in_tokens( $query );
$out_tokens = $reply->get_out_tokens();
return ( $in_tokens * $model['price']['in'] +
$out_tokens * $model['price']['out'] ) / 1000000;
}
}
return 0;
}
/**
* Retrieve the models from the OVH OpenRouter-compatible catalog.
*/
public function retrieve_models() {
// Per-request cache: build_body and the pricing helpers may all ask for
// the model list during a single query.
static $cached = null;
if ( $cached !== null ) {
return $cached;
}
$url = 'https://catalog.endpoints.ai.ovh.net/rest/v2/openrouter';
// This can run in the middle of a streaming completion, where the
// engine's http_api_curl stream handler is registered: it would hijack
// this fetch too (empty body, catalog "streamed" to nowhere) and we would
// silently fall back to the wrong model list. Detach the hooks around it.
global $wp_filter;
$saved_hooks = null;
if ( isset( $wp_filter['http_api_curl'] ) ) {
$saved_hooks = $wp_filter['http_api_curl'];
unset( $wp_filter['http_api_curl'] );
}
$response = wp_remote_get( $url, [ 'timeout' => 10 ] );
if ( $saved_hooks !== null ) {
$wp_filter['http_api_curl'] = $saved_hooks;
}
if ( is_wp_error( $response ) ) {
return $this->get_fallback_models();
}
$body = wp_remote_retrieve_body( $response );
$data = json_decode( $body, true );
if ( empty( $data ) || !is_array( $data ) ) {
return $this->get_fallback_models();
}
// Extract models from 'data' key
$models_data = isset( $data['data'] ) ? $data['data'] : $data;
if ( empty( $models_data ) || !is_array( $models_data ) ) {
return $this->get_fallback_models();
}
$models = [];
foreach ( $models_data as $model_data ) {
$model = $this->map_openrouter_model( $model_data );
if ( $model ) {
$models[] = $model;
}
}
$cached = !empty( $models ) ? $models : $this->get_fallback_models();
return $cached;
}
/**
* Map OpenRouter model data to AI Engine format.
*/
private function map_openrouter_model( $model_data ) {
if ( empty( $model_data['id'] ) ) {
return null;
}
$model_id = $model_data['id'];
$name = !empty( $model_data['name'] ) ? $model_data['name'] : $model_id;
// Extract family from model ID
$family_parts = explode( '/', $model_id );
$family = $family_parts[0];
// Base features and tags
$features = [ 'completion', 'chat' ];
$tags = [ 'core', 'chat' ];
// Check for vision support (input modality includes images)
if ( !empty( $model_data['input_modalities'] ) && is_array( $model_data['input_modalities'] ) ) {
if ( in_array( 'image', $model_data['input_modalities'] ) ) {
$features[] = 'vision';
$tags[] = 'vision';
}
}
// Map features from OpenRouter to AI Engine format
$openrouter_features = !empty( $model_data['supported_features'] ) ? $model_data['supported_features'] : [];
foreach ( $openrouter_features as $feature ) {
switch ( $feature ) {
case 'json_mode':
case 'structured_outputs':
if ( !in_array( 'json', $tags ) ) {
$tags[] = 'json';
}
break;
case 'reasoning':
if ( !in_array( 'reasoning', $tags ) ) {
$tags[] = 'reasoning';
}
break;
case 'tools':
// OVH now supports function calling via tools format
if ( !in_array( 'functions', $tags ) ) {
$tags[] = 'functions';
}
break;
}
}
// Convert pricing from per-token to per-million tokens
$price_in = 0;
$price_out = 0;
if ( !empty( $model_data['pricing']['prompt'] ) ) {
$price_in = floatval( $model_data['pricing']['prompt'] ) * 1000000;
}
if ( !empty( $model_data['pricing']['completion'] ) ) {
$price_out = floatval( $model_data['pricing']['completion'] ) * 1000000;
}
// Get context and max tokens
$max_context = !empty( $model_data['context_length'] ) ? intval( $model_data['context_length'] ) : 128000;
$max_completion = !empty( $model_data['max_output_length'] ) ? intval( $model_data['max_output_length'] ) : 4096;
return [
'model' => $model_id,
'name' => $name,
'family' => $family,
'features' => $features,
'price' => [
'in' => $price_in,
'out' => $price_out,
],
'type' => 'token',
'unit' => 1 / 1000000,
'maxCompletionTokens' => $max_completion,
'maxContextualTokens' => $max_context,
'tags' => $tags,
];
}
/**
* Get fallback models in case the catalog API is unavailable.
*/
private function get_fallback_models() {
return [
[
'model' => 'meta-llama/Llama-3.3-70B-Instruct',
'name' => 'Llama 3.3 70B Instruct',
'family' => 'meta-llama',
'features' => [ 'completion', 'chat' ],
'price' => [
'in' => 0.06,
'out' => 0.09,
],
'type' => 'token',
'unit' => 1 / 1000000,
'maxCompletionTokens' => 8192,
'maxContextualTokens' => 131072,
'tags' => [ 'core', 'chat', 'json', 'reasoning' ],
],
];
}
/**
* Get model info by model ID (for backward compatibility).
*/
private function get_ovh_models() {
return $this->retrieve_models();
}
}