Add complete Laravel LLM Gateway implementation

Core Features:
- Multi-provider support (OpenAI, Anthropic, DeepSeek, Gemini, Mistral)
- Provider service architecture with abstract base class
- Dynamic model discovery from provider APIs
- Encrypted per-user provider credentials storage

Admin Interface:
- Complete admin panel with Livewire components
- User management with CRUD operations
- API key management with testing capabilities
- Budget system with limits and reset schedules
- Usage logs with filtering and CSV export
- Model pricing management with cost calculator
- Dashboard with Chart.js visualizations

Database Schema:
- MariaDB migrations for all tables
- User provider credentials (encrypted)
- LLM request logging
- Budget tracking and rate limiting
- Model pricing configuration

API Implementation:
- OpenAI-compatible endpoints
- Budget checking middleware
- Rate limit enforcement
- Request logging jobs
- Cost calculation service

Testing:
- Unit tests for all provider services
- Provider factory tests
- Cost calculator tests

Documentation:
- Admin user seeder
- Model pricing seeder
- Configuration files

laravel-app/app/Services/LLM/Providers/GeminiProvider.php

@@ -0,0 +1,132 @@
+<?php
+
+namespace App\Services\LLM\Providers;
+
+use App\Models\ModelPricing;
+use Illuminate\Support\Facades\Cache;
+
+class GeminiProvider extends AbstractProvider
+{
+    protected string $baseUrl = 'https://generativelanguage.googleapis.com/v1beta';
+
+    protected function buildRequest(array $messages, array $options): array
+    {
+        // Gemini uses a different message format
+        $contents = [];
+        
+        foreach ($messages as $message) {
+            $role = $message['role'];
+            
+            // Gemini uses 'model' instead of 'assistant' and doesn't support 'system'
+            if ($role === 'assistant') {
+                $role = 'model';
+            } elseif ($role === 'system') {
+                // Convert system messages to user messages with context
+                $role = 'user';
+            }
+            
+            $contents[] = [
+                'role' => $role,
+                'parts' => [
+                    ['text' => $message['content']]
+                ]
+            ];
+        }
+
+        $request = [
+            'contents' => $contents,
+        ];
+
+        // Add generation config if options provided
+        $generationConfig = array_filter([
+            'temperature' => $options['temperature'] ?? null,
+            'maxOutputTokens' => $options['max_tokens'] ?? null,
+            'topP' => $options['top_p'] ?? null,
+            'stopSequences' => $options['stop'] ?? null,
+        ], fn($value) => $value !== null);
+
+        if (!empty($generationConfig)) {
+            $request['generationConfig'] = $generationConfig;
+        }
+
+        return $request;
+    }
+
+    protected function getAuthHeaders(): array
+    {
+        return [
+            'Content-Type' => 'application/json',
+        ];
+    }
+
+    public function chatCompletion(array $messages, array $options = []): array
+    {
+        $model = $options['model'] ?? 'gemini-pro';
+        $data = $this->buildRequest($messages, $options);
+        
+        // Gemini uses API key as query parameter
+        $endpoint = "/models/{$model}:generateContent?key={$this->apiKey}";
+        
+        return $this->makeRequest($endpoint, $data);
+    }
+
+    public function normalizeResponse(array $response): array
+    {
+        $candidate = $response['candidates'][0] ?? [];
+        $content = $candidate['content'] ?? [];
+        $parts = $content['parts'] ?? [];
+        
+        $textContent = '';
+        foreach ($parts as $part) {
+            $textContent .= $part['text'] ?? '';
+        }
+
+        $usageMetadata = $response['usageMetadata'] ?? [];
+
+        return [
+            'id' => null, // Gemini doesn't provide an ID
+            'model' => $response['modelVersion'] ?? null,
+            'content' => $textContent,
+            'role' => 'assistant',
+            'finish_reason' => $candidate['finishReason'] ?? null,
+            'usage' => [
+                'prompt_tokens' => $usageMetadata['promptTokenCount'] ?? 0,
+                'completion_tokens' => $usageMetadata['candidatesTokenCount'] ?? 0,
+                'total_tokens' => $usageMetadata['totalTokenCount'] ?? 0,
+            ],
+            'raw_response' => $response,
+        ];
+    }
+
+    public function calculateCost(int $promptTokens, int $completionTokens, string $model): float
+    {
+        $cacheKey = "pricing:gemini:{$model}";
+
+        $pricing = Cache::remember($cacheKey, 3600, function () use ($model) {
+            return ModelPricing::where('provider', 'gemini')
+                ->where('model', $model)
+                ->where('is_active', true)
+                ->first();
+        });
+
+        if (!$pricing) {
+            return 0.0;
+        }
+
+        $promptCost = ($promptTokens / 1_000_000) * $pricing->input_price_per_million;
+        $completionCost = ($completionTokens / 1_000_000) * $pricing->output_price_per_million;
+
+        return round($promptCost + $completionCost, 6);
+    }
+
+    public function getSupportedModels(): array
+    {
+        return [
+            'gemini-pro',
+            'gemini-pro-vision',
+            'gemini-1.5-pro',
+            'gemini-1.5-flash',
+            'gemini-ultra',
+        ];
+    }
+}