SAFi

Prerequisites

• Python 3.11+
• MySQL 8.0+ (required for JSON column support)
• Nginx or Apache (recommended for SSL/HTTPS in production)

Authentication Setup

SAFi uses OpenID Connect (OIDC) for OAuth login. If you set SAFI_LOCAL_ADMIN_EMAIL and SAFI_LOCAL_ADMIN_PASSWORD, you can skip this section entirely for local development.

Google Setup

1. Go to the Google Cloud Console. 2. Create a project and configure the OAuth consent screen. 3. Create OAuth 2.0 Client IDs (Web application). 4. Add Authorized Redirect URIs: - http://localhost:5000/api/callback (Login) - http://localhost:5000/api/auth/google/callback (Drive Integration) 5. Copy Client ID and Client Secret to .env.

Microsoft Setup

1. Go to Azure Portal → App registrations. 2. Register a new application (Accounts in any organizational directory + personal Microsoft accounts). 3. Add Redirect URIs (Web): - http://localhost:5000/api/callback/microsoft (Login) - http://localhost:5000/api/auth/microsoft/callback (OneDrive Integration) 4. Create a Client Secret under "Certificates & secrets". 5. Copy Application (client) ID and secret value to .env.

---

Setup

1. Go to Policies, create or open a policy, and generate an API key.

Manual Installation

For deploying SAFi on a bare Linux server without Docker.

Quick Start

The fastest way to run SAFi locally. Includes MySQL. No external database needed.

```bash

Developer Guide

Live Demo

safi.selfalignmentframework.com

---

2. Configure your environment

cp .env.example .env

Open .env and set:

Configuration Reference

All settings are controlled via environment variables. Copy .env.example to .env to get started.

SAFi: Self-Alignment Framework Interface

Built for IT Directors, CTOs, and Engineering VPs who need governed AI deployment without trading compliance for velocity.

At least one LLM API key (GROQ_API_KEY is free and fast to get)

API Call

POST /api/bot/process_prompt
Content-Type: application/json
X-API-KEY: sk_policy_your_key_here

{
  "user_id": "teams_user_123",
  "user_name": "John Doe",
  "message": "Can I approve this expense?",
  "conversation_id": "chat_456",
  "persona": "safi"
}

3. How to Add a Plugin (Context Injector)

Plugins run logic before the prompt reaches the LLM, useful for injecting live context (weather, user data, etc.).

1. Create the plugin -- add a file to safi_app/core/plugins/ (e.g., weather_injector.py) with an async function that returns context data.

2. Hook it in -- open safi_app/core/orchestrator.py, find process_prompt, and add your plugin to the plugin_tasks list:

   plugin_tasks = [
       # existing plugins...
       weather_injector.get_weather(user_prompt, ...)
   ]
   

1. The returned data is collected into plugin_context_data and automatically passed to the Intellect faculty.

---

How SAFi Compares

Competitor data sourced from public documentation as of May 2026. Latency and cost figures for alternatives are architecture-based estimates. Only SAFi figures are from independent adversarial testing.