Mando Workspace

Rust Cargo workspace powering the Alpiq BESS platform.

Workspace Structure

mando/
├── Cargo.toml              # Workspace root
├── Cargo.lock
├── rust-toolchain.toml     # Rust 1.88.0
├── rustfmt.toml
├── .gitlab-ci.yml          # CI/CD pipeline
├── Dockerfile              # Production image
├── container.*.Dockerfile  # Build/base images (Linux + Windows)
│
├── mando-lib/              # Core library
├── mando-lib-macro/        # Procedural macros
├── mando-bess/             # REST API service
├── py-mando/               # Python bindings (PyO3)
│
├── assets/                 # Test/demo data
├── doc/                    # Logo, diagrams
├── docs/                   # Arrow benchmarks, plans
├── lib/libduckdb/          # Bundled DuckDB 1.4.2 (Linux amd64)
├── scripts/                # Benchmarks, load tests, seed data
└── test/                   # Integration tests

Crate Dependency Graph

graph LR
    MACRO["[[mando-lib-macro]]"] --> LIB["[[mando-lib]]"]
    LIB --> BESS["[[mando-bess]]"]
    LIB --> PY["[[py-mando]]"]
    SCRT["[[mando-scrt]]"] -.-> LIB

mando-scrt

Standalone binary (not workspace member yet). Uses PyO3 auto-initialize (embeds Python) vs py-mando’s extension-module (loaded by Python). See mando-scrt.

All crates share workspace-level dependencies defined in the root Cargo.toml.

Key Workspace Dependencies

Data & Analytics

Crate	Version	Purpose
arrow	56.2.0	Columnar data (IPC/FFI)
polars	0.49.1	Lazy DataFrames, JSON, Parquet, IPC
apache-avro	0.17.0	Avro serialization

Databases

Crate	Version	Purpose
rusqlite	bundled	SQLite with r2d2 pooling
deadpool-postgres / tokio-postgres	—	PostgreSQL async
duckdb	1.4.2	DuckDB with Arrow appender
refinery	—	PostgreSQL migrations

Web & API

Crate	Version	Purpose
axum	0.8.7	Async web framework
reqwest	0.12.24	HTTP client
cynic	3.12.0	GraphQL client codegen
utoipa / utoipa-scalar	—	OpenAPI docs

Python Integration

Crate	Version	Purpose
pyo3	0.25.1	Rust ↔ Python FFI
pyo3-polars	0.22.0	Polars DataFrame bridge
pyo3-arrow	0.11.0	Arrow bridge
maturin	1.9.2	Build tool

Observability

Crate	Version	Purpose
opentelemetry	0.30.0	OTLP exporter (gRPC/tonic)
tracing	0.1.43	Structured logging
axum-otel-metrics	0.12.0	HTTP metrics

Auth & Security

Crate	Version	Purpose
jsonwebtoken	9.3.1	JWT (RS256)
rustls	0.23.35	TLS with native certs

Feature Flags

The workspace uses feature flags to control compilation:

Feature	Used By	Purpose
app	mando-lib	Enables Axum + OpenTelemetry (for mando-bess)
codegen	mando-lib	Code generation utilities
python	mando-lib	PyO3 bindings (conditional #[pyclass])
workflow	mando-lib	Flow execution engine

Architecture Patterns

Repository Pattern

Trait-based database abstraction with three backends:

• PostgreSQL — primary production store
• DuckDB — analytics/OLAP workloads
• SQLite — lightweight/local development
• Passthrough — optional caching layer

Service Layer

MandoService async trait provides uniform data access. Factory functions create backend-specific instances.

Flow Engine

Versioned workflows (v1, v2, v3) with strategy implementations for trading, auction, scheduling, etc.

Python Bridge

PyO3 extension module with abi3-py38 stable ABI. Uses pyo3-polars and pyo3-arrow for zero-copy data exchange.

Scripts

Script	Purpose
cpu_bench.sh	CPU profiling
fargate_bench.sh	AWS Fargate benchmarks
load_test.sh	HTTP load testing
seed_data.py	Test data generation
seed_scenarios.py	Scenario initialization
stress_test.py	Stress testing
scenario_stress_test.py	Scenario-level stress
bench_chart.py	Benchmark visualization