-
Notifications
You must be signed in to change notification settings - Fork 1
Architecture
Krishna Kishor Tirupati edited this page May 11, 2026
·
1 revision
PolicyAware is designed as a modular Python control plane that sits between AI applications and model or tool execution.
flowchart TD
A["Application, RAG app, or AI agent"] --> B["PolicyAware SDK or middleware"]
B --> C["Data protection engine"]
C --> D["Risk classifier"]
D --> E["Policy decision engine"]
E -->|deny| F["Blocked response"]
E -->|approval required| G["Approval workflow hook"]
E -->|allow or conditional allow| H["Model router"]
H --> I["Provider adapter"]
I --> J["Runtime evaluator"]
J --> K["Audit trace logger"]
K --> L["Application response"]
flowchart LR
subgraph Core["Core Gateway"]
G["gateway.py"]
M["models.py"]
P["policy.py"]
R["risk.py"]
D["data_protection.py"]
end
subgraph Execution["Execution"]
ROUTE["routing.py"]
PROVIDERS["providers.py"]
TOOLS["tools.py"]
APPROVALS["approvals.py"]
end
subgraph Assurance["Assurance"]
EVALS["evals.py"]
AUDIT["audit.py"]
OBS["observability.py"]
end
G --> P
G --> R
G --> D
G --> ROUTE
ROUTE --> PROVIDERS
G --> EVALS
G --> AUDIT
AUDIT --> OBS
TOOLS --> AUDIT
APPROVALS --> AUDIT
PolicyAware controls:
- Whether a request is allowed, denied, conditionally allowed, or sent for approval.
- Whether sensitive data must be redacted.
- Which model provider and model may be used.
- Whether a tool connector or action is permitted.
- How deeply a request is logged.
- Which runtime evaluations must pass.
- What compliance artifacts are retained.
PolicyAware does not replace your application, model provider, vector database, or enterprise workflow system. It provides a policy-aware boundary around them.
flowchart LR
APP["Your application"] --> PA["PolicyAware AI Gateway"]
PA --> LLM["LLM providers"]
PA --> RAG["RAG pipeline"]
PA --> MCP["MCP and tools"]
PA --> AUDIT["Audit and compliance storage"]
PA --> OBS["Metrics and traces"]
- Home
- Capabilities
- Copy-Paste Examples
- Comparison
- SEO And Distribution
- Ready-To-Use YAML
- Data Protection
- Policy Enforcement
- Gateway Orchestration
- Risk Classification
- Model Routing and Providers
- Tool Governance
- Evaluation
- Audit and Observability
- ML-Assisted Signals
- Installation
- Quick Start
- Architecture
- CLI Reference
- ML Integrations
- Provider Adapter Examples
- YAML Policy Templates