LOUIS-AI v1.0.0 – The Local, Privacy-First AI Financial Planner for Businesses & Advisors

---

🇩🇪 Suchen Sie die deutsche Dokumentation? Hier geht es zur deutschen Version (README.de.md)

(Video show the old SaaS version - free version has improved functions)

Louis.AI.-.Einfuhrung.in.die.App.KI-.Business-.Finanzplanung.1.1.mp4

---

🚀 Stop Excel Chaos & Cloud Data Leaks

Build bank-grade, institution-ready financial plans (optimized for investors and strict European funding bodies like KfW, BAFA, AVGS) 100% locally on your own infrastructure. Formerly a premium B2B SaaS—now fully open-source under GPLv3.

• Battle-Tested: In active production use by enterprises and startup consultants for over 12 months.
• Data Sovereign: Sensitive financial data, balances, and corporate strategies never leave your network.

---

⚡ Quick Start (Docker – Recommended)

Get LOUIS-AI up and running in less than 2 minutes, including a sandboxed, private local web search engine (SearXNG) out of the box:

# 1. Clone the repository and navigate into the folder
git clone <https://github.com/ren-AI-ssanceDE/LOUIS-AI> && cd louis-ai

# 2. Build and spin up the containers in the background
docker-compose up -d --build

Once deployed, open http://localhost:3000 instantly in your browser.

---

📊 Core Features

1. Comprehensive Financial Modeling & Forecasting

• Structured Workspace: Guided entry flows for corporate metadata, granular pricing matrixes, sales volumes, OpEx, CapEx (asset schedules), and dynamic financing (equity & debt schedules).
• P&L Statements (Ertragsplan): Automatic real-time formulation and aggregation of monthly and yearly revenues, materials, personnel cost, operating expenses, interest, and depreciation.
• Cash Flow Statements (Liquiditätsplan): Live cash inflow/outflow calculations computing monthly liquidity surpluses, dynamic minimum reserves, and funding bottlenecks.
• Depreciation Tables (Abschreibungsplan): Integrated with standardized asset schedules (e.g., German AfA-Tables) to automate linear depreciation values and net book values over multiple planning years.
• Scenario Comparison: Establish independent business plans (Best Case vs. Worst Case) and evaluate key performance metrics, ROI, ROS, and cash runways side-by-side via interactive D3 charts.

2. Advanced Multi-Agent Autonomous AI Team

• The Orchestrator (Louis Visionary): The command center. Parses complex user input, classifies intent (DATA_CHANGE, ANALYSIS, GENERAL), maps execution paths, and triggers agent tools.
• The Architect (Execution Assistant): Runs specialized iterative loops using the ReAct model (up to 4 iterations). Consults local RAG data, fetches anonymized web data, and solves financial mathematical equations.
• The Critic (Louis QA): Automatically intercepts data proposals, audits generated arrays for physical and mathematical consistency, filters hallucinations, and enforces core business logic.
• Visible Chain-of-Thought (CoT): Strategic thinking, SWOT indicators, and financial reasoning are displayed transparently inside a collapsible UI <thought> toggle block.

3. Zero-Direct-Write Security Model

Absolute data integrity by design: The AI cannot write directly into your workspace database. Instead, it submits structured proposals analyzed and rendered in a Visual JSON Diff Review Panel. You inspect proposed financial changes line-by-line and accept or reject them with a single click (Human-in-the-Loop).

4. Air-Gapped Local RAG Engine

Upload internal enterprise data (PDF, DocX, Excel sheets) straight into your project context. Louis AI processes files completely offline via a secure local transformer pipeline (@xenova/transformers, MiniLM) and handles lightning-fast semantic searches using a local lancedb vector store.

---

🤖 Multi-Agent Workflow Diagram

[User Query] ──> [Orchestrator] ──> Identifies Intent (DATA_CHANGE, ANALYSIS, etc.)
                          │
                          ▼
               [ReAct Loop] (Up to 4 iterations)
               ├── Chat Memory & Context Summary
               ├── Web Search (SearXNG with automated fallback checks)
               ├── Knowledge Base RAG Search (LanceDB + Local Transformers)
               └── Financial Math KPI Calculator
                          │
                          ▼
             [Louis Architect] ──> Builds Thinking & Structured Data Draft
                          │
                          ▼
             [Louis QA Critic] ──> Validates Schema & Financial Soundness
                          │
                          ▼
             [Diff Review UI]  ──> Human Manual Approval ──> Persisted to SQLite DB!

---

🛠 Tech Stack

Frontend (UI Layer)

• Vite + React (TypeScript): Ultrafast, strictly typed Single-Page Application interface.
• State Management: Zustand for unified, high-performance global states and action histories.
• Data Visualization: Recharts and D3 for professional reactive financial analytics.
• Localization: Fully localized with react-i18next (English & German pre-configured).
• Styling: Modern, modular design engineered with Tailwind CSS.

Backend (Server Layer)

• Express (Node.js): Directs API requests, secures environmental secrets, manages SMTP configurations, and handles file-system SQLite routing.

• Databases:

• SQLite (better-sqlite3) for robust, file-based persistence per project.

• LanceDB for embedded, serverless vector search.

• LLM Engine Compatibility: Out-of-the-box native integrations for:

• Ollama: Secure, locally-run open models (recommended: qwen2.5:14b or mistral).

• Google Gemini API: Utilizing the high-speed @google/genai TypeScript SDK.

• OpenAI / Claude: Global API endpoints supported.

---

⚙️ Alternative Manual Installation (Without Docker)

For detailed operating system requirements, please see INSTALL.md.

Prerequisites

• Node.js: Version 20+ (LTS recommended)
• Build Tools: Python and C++ Compilers (GCC/Clang or MS Build Tools) to compile native addons (better-sqlite3, lancedb).

# 1. Install all dependencies
npm install

# 2. Build production assets
npm run build

# 3. Fire up development mode
npm run dev

Configuration (.env)

Create a .env file in the root folder based on .env.example:

APP_URL="http://localhost:3000"
PROJECT_TOKEN="your_secure_project_token"
# GEMINI_API_KEY="AIzaSy..." # Optional, can also be supplied inside the UI admin panel

---

📄 License & Open-Source Terms

This software is licensed under the GNU General Public License v3 (GPLv3).

• Free as in Freedom: Anyone is permitted to run, audit, modify, and redistribute LOUIS-AI.
• Copyleft Obligations: If you distribute modified versions of this software, or host modified versions of it as a paid online service (SaaS), you MUST publish your complete modified source code under the same GPLv3 license. Closed-source commercial forks are strictly prohibited.

---

💼 Enterprise Support & Customization

LOUIS-AI is and will always remain 100% free. However, if you require professional deployment assistance, integration into legacy ERP/banking software, or custom feature branches tailored to your corporate workflow, the official core maintainer is available for hire:

ren-AI-ssance (Owner: Stefan Tusk) Pragmatic AI Solutions & Strategic Management Consulting.

• Website: https://ren-ai-ssance.de
• Email: service@ren-ai-ssance.de
• Contact: Stefan Tusk

---

Built with passion for data-sovereign corporate financial intelligence – www.ren-ai-ssance.de