composer

Wink Composer provides a framework to build and run composable streaming analytics pipelines, exposed as an MCP Server. It centers on a Flow language that lets you declaratively stitch together small, focused building blocks (nodes) to process live streams, extract features, detect events, and persist insights. The server surfaces AI-ready insights by creating structured, contextualized results from streaming data, enabling retrieval over insights rather than raw data and supporting edge-to-cloud deployments with consistent code paths across environments. Tools and capabilities highlighted in the project include building-block nodes for signal conditioning, feature extraction, detection, flow control, emission, and data quality, all orchestrated through a single Flow pipeline that can scale from individual pipelines to fleet-wide analytics. With connected storage (e.g., QuestDB) and input/output integrations, Composer enables end-to-end streaming intelligence, from ingestion to persistent insights, while maintaining isolated state per sensor, user, or session for resilience and scalability.

Prerequisites:

• Node.js (LTS) installed on your system
• npm or pnpm to install dependencies
• Git to clone the repository (optional if downloading as a package)

Step-by-step installation:

1. Clone the repository or download the source: git clone https://github.com/winkjs/composer.git cd composer

2. Install dependencies: npm install

3. Build or prepare the server if the project provides a build step (adjust as needed): npm run build

4. Start the MCP Server locally (adjust script if your package defines a different start command): npm run start

5. Verify the server is running by checking logs or hitting the health endpoint (if provided by the project). If you use a custom entry point, ensure server.js (or your specified entry) is present and accessible.

Notes:

• If your project uses environment variables for configuration (e.g., database URLs, broker addresses), create a .env file or export them in your shell before starting the server. The MCP config can also reference these env vars as needed by your deployment.

Tips and considerations:

• Environment variables: configure your data storage (QuestDB), message brokers (MQTT, Kafka), and AI/LLM integrations via environment variables as documented by the project.
• Flow language: design modular pipelines using the 6 categories of blocks (Signal Conditioning, Feature Extraction, Detection, Flow Control, Emission, Data Quality) to keep pipelines readable and maintainable.
• Observability: leverage built-in logging, metrics, and error handling to diagnose issues quickly in streaming scenarios.
• Deployment: this MCP server is intended to run consistently across edge to cloud; maintain consistent dependencies and entry points to minimize drift between environments.
• Common issues: ensure correct broker URLs, storage endpoints, and network access; verify that storage like QuestDB is reachable and listening on expected ports; confirm that the Flow steps reference valid node names and that any required prerequisites (e.g., QuestDB, MQTT broker) are up before starting.
• Configuration: use the provided mcp_config block to wire the server into your MCP orchestration layer, and tailor env vars to your deployment environment.