enterprise-ai-office

Enterprise AI Office is a comprehensive system blueprint and installation guide for constructing a self-hosted, governed enterprise AI workspace. The project focuses on integrating three core components: **Open WebUI** as the user interface, **Hermes Agent** as the runtime orchestration layer, and **WeKnora** as the authoritative company knowledge base. The repository emphasizes security, governance, and reproducibility, particularly through agent-readable contracts that allow AI engineering agents to reconstruct and validate the system from scratch. Key features include: - **Core Employee Path**: A validated workflow where employees interact with a general assistant via Open WebUI, which queries Hermes and retrieves grounded answers from WeKnora. - **Governed Communication (v2)**: A specialized capability for email handling that enforces strict governance. This includes domain models for drafts, approvals, and send actions, ensuring that external communications require deterministic human review before sending. It supports reconciliation for ambiguous outcomes and maintains immutable audit trails in a lightweight SQLite backend. - **Least-Privilege Operations**: An ARMOR reference deployment profile that restricts agent capabilities to specific business tasks (e.g., content preparation) without granting broad filesystem or shell access. - **Installation & Deployment Blueprints**: Detailed documentation (`DEPLOY.md`, `AGENTS.md`) and machine-readable manifests (`eao-manifest.yaml`) that define installation stages, protected inputs, identity propagation, and rollback procedures. - **Deferred Scope**: The project intentionally excludes CRM, ERP, PIM, and complex observability stacks in its baseline to maintain simplicity and security, focusing instead on proving the core governance pattern. The system is designed to be extensible via optional capabilities such as coding agents, messaging integrations, and specialist profiles, all configurable through YAML-based capability registries.