AXIOM v0.2.0-beta.7 — Optimizer, macOS CLI and beta updates

AXIOM AI v0.2.0-beta.7

Build AI. Own AI.

AXIOM AI is an AI engineering platform focused on model management,
optimization workflows, dataset preparation, training planning, evaluation,
and local runtime tooling.

Release status: Beta / pre-release.

This release is intended for testing and feedback. Features and interfaces
may change before a stable release. Review the repository's security
documentation before using AXIOM with sensitive data or untrusted models.

Highlights

Model optimization

• Introduces the AXIOM optimizer workflow for inspecting models and preparing
  runtime-focused bundles.
• Adds agent profiles for coding, reasoning, research, general assistance,
  automation, mathematics, writing, and multilingual tasks.
• Supports system-aware optimization planning and configurable performance
  targets.
• Treats performance targets as goals, not guarantees; actual throughput
  depends on the model, hardware, runtime, context length, and configuration.

Model policy auditing

• Adds a model policy audit workflow to identify policy- and safety-related
  indicators in model files and configurations.
• The audit is informational. It does not guarantee that a model is safe,
  and it does not automatically remove model safeguards.

Security and path handling

• Adds model-path validation and boundary checks to optimization-related
  workflows.
• Continues hardening API routes and filesystem operations involving model
  paths.
• Adds administrator-storage handling in the frontend.

Security scanning is an ongoing process. Users should check the repository's
Security tab for current CodeQL findings and should not interpret this beta
release as proof that all security findings have been resolved.

Frontend and usability

• Improves the AXIOM Control Center's optimizer and quantization interfaces.
• Includes automated browser coverage for first-boot authentication and
  invalid-login handling.
• Continues work on the dark engineering-dashboard interface.

Documentation

• Updates release documentation and the project roadmap.
• Documents the beta's scope, limitations, and validation expectations.

Validation

The following checks have passed in the development environment during
release preparation:

• Python test suite: 23 tests passed.
• Playwright end-to-end suite: 2 tests passed.
• ESLint.
• TypeScript compilation and Vite production build.
• Python bytecode compilation.
• Ruff checks.
• Git whitespace/error checks.

These results describe the tested development state; they do not guarantee
that every supported operating system, model, accelerator, or deployment
configuration has been tested.

Known limitations

• Beta software may contain bugs and breaking changes.
• Model performance and memory use vary significantly by hardware and model.
• Optimization plans and performance targets are not guarantees of measured
  inference speed.
• Policy auditing is not a complete safety assessment.
• Security alerts may be updated as new scans run. Check the repository's
  current CodeQL results before deployment.
• Review the project's installation instructions for platform-specific
  requirements before running workloads.

Upgrade and installation

Use the installation and setup instructions in the repository README.
Back up project configuration and any important local data before upgrading.
Do not assume that beta versions are interchangeable without checking the
release-specific instructions.

Feedback and bug reports

Please report reproducible bugs through GitHub Issues. Include your operating
system, Python and Node.js versions, the command or workflow used, relevant
logs, and steps to reproduce the issue. Remove passwords, access tokens,
private model data, and other secrets before sharing logs.

Contributors

Thank you to everyone testing AXIOM and reporting bugs, usability problems,
and improvement ideas.

---

AXIOM AI — Build AI. Own AI.