v4.2.1 Documentation

Introduction to SIFRA

SIFRA (Synthetic Intelligence for Real-time Analysis) is an autonomous data scientist engine designed to bypass the latency of human analysis. Unlike traditional LLM wrappers, SIFRA utilizes HDP-FusionNet, a proprietary vector-based intent engine, to perform hard mathematical computations, forecasting, and anomaly detection instantly.

Deterministic

Results are mathematically verifiable, not hallucinated.

Zero-Config

Auto-detects schema, seasonality, and correlations.

Core Architecture

The system is composed of three primary nodes operating in parallel:

HDP-FusionNet

High-Dimensional Processing layer. It ingests raw CSV/JSON streams and converts them into normalized vector tensors for analysis.

HDS-Unity Matrix

The central state machine. It orchestrates the flow between the forecasting engine, the anomaly detector, and the visualization renderer.

Quick Start Guide

Follow this protocol to initiate your first analysis run.

  1. 1

    Upload Dataset

    Navigate to the Command Center and upload a CSV file. Ensure the first row contains headers.

    Date,Revenue,Cost 2024-01-01,1500,800 2024-01-02,1800,900
  2. 2

    Select Operation

    Choose "Deep Analysis" or "Forecasting" from the Input Parameters panel.

  3. 3

    Initialize Engine

    Click the "Initialize Engine" button. The system log will display real-time processing events.

API Reference

Titan Tier Required

Direct API access is restricted to Titan subscribers. Recruit and Commander users must use the web interface.

Authentication

Include your Secret Key in the header of every request.

# Example Authentication Header Authorization: Bearer sk_live_9482...

Prediction Endpoint

Send data for real-time inference using a pre-trained model.

POST /v1/predict
curl -X POST https://api.sifra.ai/v1/predict \ -H "Authorization: Bearer sk_live_..." \ -H "Content-Type: application/json" \ -d '{ "model_id": "model_alpha_01", "features": [150, 2.5, "Q4"] }'