🔍 Spryker Project Semantic Search Tool

The Spryker Project Semantic Search Tool enables intelligent, context-aware search across your Spryker project using Large Language Model (LLM)-based embeddings and Chroma DB for efficient indexing and retrieval.

🧩 Problem

Spryker applications consist of a vast and growing number of module APIs, including Facades, Clients, and various Plugin interfaces used as extension points. Due to this complexity, identifying the right module or method for a specific task can be time-consuming and error-prone, especially for new developers or when working across multiple feature sets.

✨ Features

• Indexes key module APIs and configurations, including:

  • Facade interfaces
  • Client interfaces
  • Service interfaces
  • Plugin interfaces and plugin classes
  • Config classes of modules

• Semantic understanding of code:

  • Uses class and method names (module APIs)
  • Incorporates method doc blocks for deeper intent and specification analysis
  • Does not use full file code

• Security Measures:

  • Prevents exposure of sensitive project data by operating in local mode.
  • Extracts only semantic information from module APIs, without analysing underlying code implementations.

• Efficient navigation and output:

  • Presents results in chunked format for readability
  • Each result includes a link to the source file and line number for quick access
  • Various AI modes for better usability

🧠 How It Works

1. Embeddings Generation: Creates vector embeddings for indexed elements using a Large Language Model (LLM).
2. Indexing: Stores embeddings in Chroma DB for fast semantic search.
3. User Query: Accepts natural language input.
4. Matching & Ranking: Finds semantically relevant matches across the project and summarises with AI.
5. Result Presentation: Displays readable chunks with direct links to source files.

📦 Use Cases

• Understand large or unfamiliar Spryker codebases faster.
• Discover relevant module APIs and plugins by use case, not just name.
• Accelerate onboarding for new developers and cross-team collaboration.

---

💡 Tip: Works best when combined with up-to-date PHPDoc across modules for optimal semantic accuracy.

🔍 Filtering

The tool supports advanced Machine Learning-based filtering detection. To use this feature, simply mention it in your query — no additional setup is required beyond providing a suitable Facade, Configuration, Client, Plugin etc. Additionally, the tool is capable of understanding extended filter definitions configured in \Spryker\Config::getDataTypeTrainData, which you can override to customize detection behavior for your specific use case.

🧪 Examples:

Native mode

Native plus AI mode

📦 Prerequisites

Ensure you have the following installed on your machine:

• Docker
• Docker Compose
• Composer
• Bash shell

---

🛠 Installation & Setup

1. Clone the repository into the project root directory:

   git clone git@github.com:spryker-community/project-semantic-search.git &&
   echo "/project-semantic-search/" >> .git/info/exclude &&
   cd project-semantic-search &&
   cp php/.env.example php/.env

2. Configure environment in project-semantic-search/php/.env

3. Run the installer script:

bash install

This will:

* Start Docker containers
* Install dependencies via Composer
* Pull ollama embedding model nomic-embed-text
* Index the project (takes 5–20 minutes)
* Launch the interactive CLI tool

---

📂 Project Structure

├── spryker-project/
│      ├── src/
│      │    └── Pyz/
│      ├── ...
│      └── project-semantic-search/
│           ├── php/
│           │    └── bin/
│           │         └── sprykeye
│           ├── docker-compose.yml
│           ├── php.ini
│           ├── install
│           ├── run
│           └── readme.md

---

🧪 Usage

After setup, you can launch the search tool anytime by running:

docker exec -it php bash -c "bin/sprykeye project:search"

or

bash run

MCP server makes the tool compatible with various AI agents to extend context with Spryker Project context

Tools:

• Claude desktop
• JetBrains AI Assistant plugin
• Qodo Gen JetBrains plugin
• Windsurf
• Cursor
• Cursor
• etc.

How to configure?

Example:

📈 Roadmap

• Controller classes indexing (public methods);
• Dependency provider's extension points;
• Filtering by Module name;
• AI chat mode - discuss with AI results or better query;
• AI: search with remote AI agent (Gemini, OpenAI, etc.);

📄 License

MIT or your preferred license.

---

👥 Authors

• Vitalii Ivanov