Monolith is an alpha-stage local AI workbench for experimenting with local LLMs, evaluation prompts, model profiles, and workstation-oriented workflows.

Monolith is intended to run locally on a trusted machine.

The repository should not contain:

• API keys
• tokens
• passwords
• private SSH keys
• .env files
• real local model configuration files
• SQLite databases
• raw logs
• model binaries
• screenshots with private data

Real local configuration files such as configs/models.yaml and configs/test_profiles.yaml are intentionally ignored by Git.

Use the sanitized examples under configs/, such as configs/models.example.yaml, as templates.

Monolith includes early Agent Lab and evaluation scaffolding.

At this stage:

• agent workflows should remain bounded and review-first
• command execution should not be automatic
• file mutation should require explicit review
• arbitrary workspace paths should not be trusted
• Docker socket mounts, SSH key mounts, and broad host mounts should be treated as high-risk

For now, report security concerns privately to the repository owner.

Do not open a public issue containing secrets, tokens, local paths, screenshots, or sensitive logs.