← Back to Chat
⚙️

Admin Governance Panel

💬 Chat 📊 Diagram 🚨 Security Dashboard
🛡️

Centralized Policy Engine

Configure user token budgets, LLM provider credentials, agent tool rules and role privileges. All changes hot-reload in runtime with zero restarts.

Policy Store: SQLite ● Live Hot-Reload Active

💰 User Budget Caps & Spend Governance

Enforce hard dollar caps on LLM spend. Requests exceeding caps are rejected instantly (<1ms, $0) without downstream token burn.

User Role Current Total Spend Hard Budget Cap (USD) Update Cap
Loading user budget limits...

🔑 External Providers & API Tokens

Configure upstream AI provider credentials stored securely in SQLite.

🧰 Agent Tool Rules (Hybrid Tool Gate)

Every tool call: deny rules (final, even for admin) → argument DLP scan → allow rules → gray zone: AI tool guard + role privileges. Tool patterns match the whole tool name; ALLOW argument patterns must match the whole value, DENY patterns match anywhere. Changes apply instantly.

#ActionTool ArgumentPatternRole Description
Test a tool call (dry run, nothing executes)

👥 Roles & Privileges

Toggle a privilege to grant or revoke it for the whole role. Applies to the next tool call.

⏸ Tool Call Approval Queue

Paused agent runs resume as soon as their last pending call is decided.