# LLM Council – Personal SAP AI Intelligence System

A system where five AI models analyze every SAP question and a chairman model synthesizes the best answer.

All components run locally and use free AI APIs.

---

## Overview

LLM Council is a personal AI advisory system developed to improve the quality of answers for complex SAP questions.

Instead of relying on a single AI model, the system creates a structured evaluation process:

**Stage 1 – Multi-Model Response**

The user's question is sent to five different AI models simultaneously.

**Stage 2 – Peer Review**

Each AI model reviews and ranks the answers generated by the other models.

**Stage 3 – Chairman Synthesis**

A chairman model analyzes all responses and reviews, then produces a final consolidated answer.

The result simulates a panel of AI experts debating the question before producing a final recommendation.

---

## Features

Five AI council members

* Llama 3.3 70B

* Llama 3.1 8B

* Gemma 2 9B

* Mixtral 8x7B

* Llama 3.2 11B

Chairman AI

* Llama 3.3 70B synthesizes the final response

Capabilities

* Image upload for SAP screenshot analysis

* Document upload for PDF and Word functional specifications

* Voice input using Whisper transcription

* Multi-turn conversations with context retention

* Automatic local conversation storage as JSON files

* Runs locally on the user's machine

* Uses the Grok API free tier

---

## Technology Stack

Frontend

React 18, Vite, ReactMarkdown

Backend

Python, FastAPI, Uvicorn

AI Models

Llama, Gemma, Mixtral through the grok API

Voice Processing

Whisper Large V3 via grok

Vision Processing

Llama 4 Scout via grok

Package Management

uv

---

## Quick Start

### Prerequisites

* Python 3.10 or higher

* Node.js 18 or higher

* grok API key from [https://console.grok.com](https://console.grok.com)

---

### Installation

Clone the repository

git clone https://github.com/YOUR\_USERNAME/llm-council-sap.git

cd llm-council-sap

Create environment file

echo "grok\_API\_KEY=your\_grok\_key\_here" > .env

Install backend dependencies

python -m uv sync

Install frontend dependencies

cd frontend

npm install

cd ..

---

### Running the Application

Start the backend

python -m uv run python -m backend.main

Start the frontend

cd frontend

npm run dev

Open the application in the browser

http://localhost:5173

---

## SAP Use Cases

Error analysis

Upload SAP dump screenshots for visual diagnosis.

Configuration guidance

Request step-by-step configuration instructions.

Functional specification review

Upload functional specification documents and identify potential gaps.

ABAP code review

Paste ABAP code to receive analysis and improvement suggestions.

Process design

Ask for best practice recommendations for SAP business processes.

Troubleshooting

Describe system issues and receive multi-model analysis and solutions.

---

## Motivation

As an SAP Functional Consultant, reliable answers to complex SAP questions are essential. Individual AI models can sometimes produce incomplete or incorrect responses.

This project uses a structured multi-model review process to simulate collective expert reasoning, improving reliability and depth of analysis.

---

## Screenshots

Screenshots will be added in future updates.

---

## Credits

Concept inspired by the LLM Council idea developed by Andrej Karpathy.

This implementation extends the concept with multimodal capabilities, a professional user interface, and SAP-focused workflows.

Powered by the Grok API.

---

## License

MIT License.

Free to use, modify, and distribute.

LLM-Council-SAP

Personal AI Council - 5 LLM models debate SAP questions. Built with FastAPI, React, Grok API. FREE to run locally.

3111aee40c39bc92e0fe609a434819af34fc38fc

---

Built by Abhishek Singh — SAP Functional Consultant