Architecture

Overview

Briefcase AI is a layered system: a Rust core provides high-performance snapshot management, PyO3 bindings expose it to Python, and framework integrations connect to popular AI libraries.

graph TD
    A[Framework Integrations] --> B[Python Package]
    B --> C[PyO3 Bindings]
    C --> D[Rust Core]
    D --> E[Storage Backends]

Rust Core

The Rust core handles:

• DecisionSnapshot creation and serialization
• Storage backend abstraction
• Replay engine
• Drift detection algorithms

PyO3 Bindings

PyO3 bridges Rust and Python. The bindings are built with maturin and provide zero-copy access to Rust data structures where possible.

Python Package

The Python layer adds:

• @capture decorator
• Extensible framework protocols (guardrails, exporters, routers, events)
• Configuration and lazy imports

Decision Flow

sequenceDiagram
    participant App
    participant Capture as capture decorator
    participant Core as Rust Core
    participant Storage

    App->>Capture: call function
    Capture->>Core: create DecisionSnapshot
    Core->>Storage: persist snapshot
    Storage-->>Core: snapshot ID
    Core-->>Capture: snapshot reference
    Capture-->>App: function result