Limits And Guarantees¶
What We Can Claim¶
This project should earn trust by being specific about what is guaranteed and what is merely current behavior.
flowchart LR
A[Documented contracts] --> D[Claims we can stand behind]
B[Current tests and parity checks] --> D
C[Maintainer review] --> D This diagram shows where confidence is allowed to come from. Bijux treats contracts, live tests, and current review evidence as the inputs that justify a claim, instead of relying on reputation or broad language.
flowchart TD
A[Request stronger guarantee] --> B{Is it documented?}
B -->|Yes| C[Treat as a contract]
B -->|No| D[Treat as current behavior only]
D --> E[Check tests and release notes before relying on it] This decision path explains how to classify a behavior. If the answer is not in the documented contract set, the safe reading is that it remains current behavior only and should be checked before you build automation around it.
Where Bijux Fits Well¶
Bijux is a strong fit when you need:
- deterministic behavior in scripts and CI
- a command surface that will grow over time
- plugin support without pretending plugins are sandboxed
- explicit output and compatibility review
Bijux is a weak fit when you need:
- a tiny throwaway CLI with almost no policy surface
- Windows support today
- hard isolation between host and plugin code
Current Hard Limits¶
- Windows is not a supported platform
- plugins are not sandboxed
- not every internal crate API is a public compatibility commitment
- release quality is judged from live tests and maintainer checks, not from checked-in artifact snapshots
The Windows limit is about the supported host contract, not about pretending no Windows-aware code exists. The current runtime can resolve Windows-style binary names and state paths, but the shipped plugin scaffold and completion guidance still assume Linux/macOS host behavior.
Current Guarantees¶
- the repository treats the Rust runtime as the behavioral source of truth
- the Python package is checked against the current runtime and the latest stable PyPI compatibility baseline configured for the parity suite
- generated artifacts under
artifacts/are disposable and not part of the repo contract
How To Judge Maturity¶
Use these sources, in this order:
Closing Rule¶
If a claim is not grounded in current docs, tests, or release evidence, do not treat it as guaranteed.