How owoeye-babatunde/robotrade works
Main Components
• Front/Client Side (if applicable): Although not explicitly a “frontend”, the inference pipeline provides a websocket-based API which acts as the consumer interface.
• Data Ingestion & Trade Streaming: The “trades” service connects to external data sources (e.g., Kraken Websocket API) for real-time trade data.
• Data Transformation & Feature Engineering: Services like “candles” (for converting trade streams to OHLCV data) and “technical-indicators” (for computing indicators with Ta-Lib) process the raw trade data.
• News & Signal Processing: The “news” and “news-signal” services fetch, preprocess, and transform news data – integrating NLP processing – to generate trade signals.
• Feature Store Integration: The “to-feature-store” service pushes engineered features from traded data and news signals into the Feature Store.
• Machine Learning & Prediction:
• “price-predictor” is responsible for model training and prediction based on the feature store data.
• “prediction-api”, developed in Rust, offers a high-performance API endpoint for serving predictions (likely via websockets).
• Supporting Infrastructure: Docker, Docker Compose, and automated CI/CD workflows (GitHub Actions, pre-commit hooks, Ruff configuration) are part of the overall delivery pipeline.
Representing Interactions and Data Flows
• That both the training pipeline and the inference pipeline are downstream consumers of these features.
• Relationship from the “price-predictor” to “prediction-api” where the trained model is deployed, and the prediction API serves real-time prediction requests.
Architectural Patterns and Design Principles
• Microservices approach by isolating each major function (trades, candles, technical-indicators, news, news-signal, to-feature-store, price-predictor, prediction-api) in separate service containers.
• Usage of stateful stream processing and containerization to ensure scalability and robust deployment.
• That critical high-performance components may use Rust, while most services are implemented in Python.
• Point out integration with Kafka/Redpanda as the backbone for real-time data streaming.
Incorporate Specific Technologies
• Integration with tools such as Quix Streams for stateful processing and feature store management.
• Mention deployment orchestration via Docker Compose and the configuration management (environment settings and Dockerfiles).