Bayesian Optimization sidecar for CryptoBot — a FastAPI + Optuna TPE sampler service that provides intelligent parameter tuning via REST API.
Despite the name, this is not a backup of CryptoBot. It contains:
- Optuna Bayesian Optimizer (Python/FastAPI): A sidecar service that runs TPE (Tree-structured Parzen Estimator) trials to optimize trading strategy parameters, exposed as a REST API
- Launcher (
launch.py): Orchestrates startup of both the .NET CryptoBot engine and the Python optimizer sidecar - AI Governance Documents: Multi-AI collaboration protocols and handoff records from the development process
| Component | Technology |
|---|---|
| Optimizer | Python, FastAPI, Optuna TPE, Pydantic, Uvicorn |
| Launcher | Python (subprocess orchestration) |
| Target System | .NET CryptoBot (git submodule, not included in this repo) |
launch.py
├── starts CryptoBot (.NET) ← git submodule
└── starts Optimizer (FastAPI) ← localhost-only sidecar
└── Optuna TPE sampler
└── REST API for parameter suggestions
- Sidecar pattern: The optimizer runs as a separate process, communicating via REST. This keeps Python ML dependencies isolated from the .NET trading engine.
- Loopback only: The FastAPI sidecar binds to localhost — no external network exposure by design.
- Pydantic schemas: Request/response models are fully typed for parameter ranges and trial results.
The CryptoBot/ directory is a git submodule pointing to the main CryptoBot repository. Clone with --recurse-submodules to include it.