Weave Code
Integration bundle for Symfony AI components
Frequently asked questions about Ai Bundle
- Can I use Symfony AI Bundle in Laravel? If so, how does it integrate?
- No, this bundle is designed exclusively for Symfony 6.4+ applications. While Laravel shares some PHP/Symfony ecosystem components, this bundle relies on Symfony’s Dependency Injection, attributes (PHP 8.2+), and framework-specific services like ScopingHttpClient. For Laravel, consider alternatives like Laravel AI SDKs or custom integrations with OpenAI’s API.
- What Laravel alternatives exist for AI integration with similar abstraction?
- For Laravel, explore packages like `laravel-ai` (official), `spatie/laravel-ai` (community-driven), or `php-ai/php-ai` for provider-agnostic abstractions. These focus on Laravel’s service container and Eloquent, while Symfony AI Bundle prioritizes Symfony’s ecosystem. Migration would require rewriting DI and security logic.
- How do I configure OpenAI as a platform in Symfony AI Bundle?
- Add the OpenAI platform to your `config/packages/symfony_ai.yaml` under `ai.platform.openai` with your API key and model settings. Example: `api_key: '%env(OPENAI_API_KEY)%'`. The bundle auto-registers the service via Symfony’s DI, so no manual `services.yaml` setup is needed unless customizing defaults.
- Does Symfony AI Bundle support vector stores like Pinecone or MongoDB?
- Yes, the bundle includes pre-built bridges for Pinecone, Qdrant, and MongoDB via `StoreFactory`. For MongoDB, ensure you install `mongodb/mongodb` and configure the `ai.store.mongodb` service in YAML. Local stores like SQLite-Vec are also supported, but cloud providers may require additional API key setup.
- How do I secure AI tools with Symfony’s security system?
- Use the `#[IsGrantedTool]` attribute on tool methods to enforce role-based access (e.g., `#[IsGrantedTool('ROLE_ADMIN')]`). The bundle integrates with Symfony’s Security component, so existing voter logic applies. For Laravel, you’d need to manually implement middleware or gates around AI service calls.
- What’s the performance impact of using SQLite-Vec for vector storage?
- SQLite-Vec is lightweight for development but may struggle with high-throughput applications due to disk I/O bottlenecks. For production, prefer cloud-based stores like Pinecone or Qdrant. Benchmark with your dataset size—test queries with 10K+ vectors to identify latency spikes.
- How do I test AI agents in Symfony with mocked platforms?
- Use Symfony’s `HttpClientMock` to intercept API calls to OpenAI/Azure. For agent testing, mock the `ModelClientInterface` and inject it via DI. Stateful agents (e.g., chat history) require resetting mocks between tests or using fixtures to avoid flakiness. The bundle doesn’t provide built-in test utilities.
- What happens if I upgrade Symfony from 6.4 to 7.0 with this bundle?
- The bundle is designed for Symfony 6.4+ and should work with 7.0, but test thoroughly. Breaking changes may occur in experimental features (e.g., `VecStore`). Check the [Symfony AI changelog](https://github.com/symfony/ai) for version-specific notes. Always update dependencies via Composer and validate integration points like `ScopingHttpClient`.
- Can I customize the input/output processing for AI responses?
- Yes, implement `InputProcessorInterface` or `OutputProcessorInterface` to transform prompts/responses. For example, sanitize inputs with `HtmlPurifier` or format JSON outputs. Register your processor as a service and tag it with `ai.processor`. The bundle’s `DefaultProcessor` handles basic text processing.
- Are there known issues with streaming responses (DeltaInterface) in production?
- Streaming via `DeltaInterface` is experimental and may introduce edge cases like partial responses or connection timeouts. Monitor token limits and implement retry logic for failed streams. For Laravel, consider using OpenAI’s native streaming API directly if stability is critical.
Weaver
How can I help you explore Laravel packages today?