Merge pull request #36 from investbrainapp/feat-use-openai-compatible-endpoints

feat: adds ollama support

README.md

@@ -20,7 +20,7 @@ Investbrain is a smart open-source investment tracker that helps you manage, tra
 
 ## Under the hood
 
-Investbrain is a Laravel PHP web application that leverages Livewire and Tailwind for its frontend. Most databases should work, including MySQL and SQLite. Out of the box, we feature three market data providers: [Yahoo Finance](https://finance.yahoo.com/), [Finnhub](https://finnhub.io/pricing-stock-api-market-data), and [Alpha Vantage](https://www.alphavantage.co/support/). But we also offer an extensible market data provider interface for intrepid developers to create their own! We also offer an integration with OpenAI's LLMs for our ["chat with your holdings"](#chat-with-your-holdings) capability. Finally, of course we have robust support for i18n, a11y, and dark mode. 
+Investbrain is a Laravel PHP web application that leverages Livewire and Tailwind for its frontend. Most databases should work, including MySQL and SQLite. Out of the box, we feature three market data providers: [Yahoo Finance](https://finance.yahoo.com/), [Finnhub](https://finnhub.io/pricing-stock-api-market-data), and [Alpha Vantage](https://www.alphavantage.co/support/). But we also offer an extensible market data provider interface for intrepid developers to create their own! We also offer an integration with OpenAI for our ["chat with your holdings"](#chat-with-your-holdings) capability. Finally, of course we have robust support for i18n, a11y, and dark mode. 
 
 ## Self hosting
 
@@ -58,7 +58,9 @@ Investbrain offers an AI powered chat assistant that is grounded on *your* inves
 
 When self-hosting, you can enable the chat assistant by configuring your OpenAI Secret Key and Organization ID in your [.env](https://github.com/investbrainapp/investbrain/blob/main/.env.example) file. Navigate to OpenAI to [create your keys](https://platform.openai.com/api-keys).
 
-Always keep in mind the limitations of large language models. When in doubt, consult a licensed investment advisor. 
+If you are self-hosting your own large language models ("LLMs") that expose an OpenAI compatible API (e.g. [Ollama](https://ollama.com/blog/openai-compatibility)), you can update the `OPENAI_BASE_URI` configuration to your self-hosted instance. Ensure you also update the `OPENAI_MODEL` to an available model.
+
+Always keep in mind the limitations of LLMs. When in doubt, consult a licensed investment advisor. 
 
 ## Market data providers
 
@@ -133,11 +135,12 @@ There are several optional configurations available when installing using the re
 | AI_CHAT_ENABLED | Whether to enable AI chat features | `false` |
 | OPENAI_API_KEY | OpenAI secret key (required for AI chat) | `null` |
 | OPENAI_ORGANIZATION | OpenAI org id (required for AI chat) | `null` |
+| OPENAI_MODEL | The selected LLM used for AI chat | gpt-4o |
+| OPENAI_BASE_URI | The URI for your self-hosted LLM | api.openai.com/v1 |
 | DAILY_CHANGE_TIME | The time of day to capture daily change | 23:00 |
 | REGISTRATION_ENABLED | Whether to enable registration of new users | `true` |
 
 
-
 > Note: These options affect the [docker-compose.yml](https://github.com/investbrainapp/investbrain/blob/main/docker-compose.yml) file, so if you decide to make any changes to these default configurations, you'll have to restart the Docker containers before your changes take effect.
 
 ## Updating

composer.json

@@ -16,7 +16,7 @@
         "livewire/livewire": "^3.5",
         "livewire/volt": "^1.6",
         "maatwebsite/excel": "^3.1",
-        "openai-php/laravel": "^0.10.2",
+        "openai-php/client": "^0.10.3",
         "predis/predis": "^2.2",
         "robsontenorio/mary": "^1.35",
         "scheb/yahoo-finance-api": "^4.11",

config/openai.php

@@ -27,5 +27,6 @@ return [
     'request_timeout' => env('OPENAI_REQUEST_TIMEOUT', 30),
 
     //
+    'base_uri' => env('OPENAI_BASE_URI', 'api.openai.com/v1'),
     'model' => env('OPENAI_MODEL', 'gpt-4o'),
 ];

resources/views/livewire/ai-chat-window.blade.php

@@ -5,7 +5,7 @@ use App\Models\AiChat;
 use App\Models\Holding;
 use Illuminate\Database\Eloquent\Model;
 use Livewire\Volt\Component;
-use OpenAI\Laravel\Facades\OpenAI;
+use OpenAI;
 use OpenAI\Responses\StreamResponse;
 
 new class extends Component {
@@ -67,7 +67,9 @@ new class extends Component {
     {
     
         try {
-            $stream = OpenAI::chat()->createStreamed([
+            $client = $this->createOpenAiClient();
+            
+            $stream = $client->chat()->createStreamed([
                 'model' => config('openai.model'),
                 'messages' => [
                     ['role' => 'system', 'content' => "Today's date is "
@@ -104,7 +106,9 @@ new class extends Component {
     public function generateSuggestedPrompts(): void
     {
         try {
-            $suggested_prompts = OpenAI::chat()->create([
+            $client = $this->createOpenAiClient();
+
+            $suggested_prompts = $client->chat()->create([
                 'model' => config('openai.model'),
                 'response_format' => [
                     'type' => 'json_schema',
@@ -192,6 +196,25 @@ new class extends Component {
         return false;
     }
 
+    private function createOpenAiClient()
+    {
+        $apiKey = config('openai.api_key');
+        $organization = config('openai.organization');
+        $baseUri = config('openai.base_uri');
+
+        if (! is_string($apiKey) || ($organization !== null && ! is_string($organization))) {
+            throw new \InvalidArgumentException('The OpenAI API Key is missing. Please publish the [openai.php] configuration file and set the [api_key].');
+        }
+
+        return OpenAI::factory()
+            ->withApiKey($apiKey)
+            ->withOrganization($organization)
+            ->withHttpHeader('OpenAI-Beta', 'assistants=v2')
+            ->withHttpClient(new \GuzzleHttp\Client(['timeout' => config('openai.request_timeout', 30)]))
+            ->withBaseUri($baseUri)
+            ->make();
+    }
+
 }; ?>
 
 <div