📺 Watch the full tutorial on YouTube

🧹 Xberg Data Launderer

Automating My Boring Office Job with 50 Lines of Python — An enterprise-grade, local AI pipeline that transforms messy spreadsheets, broken CSVs, and Word documents into clean, structured data and an interactive single-file HTML dashboard.

---

💡 Why This Project Exists

Office workers and data analysts spend countless hours cleaning broken .xlsx spreadsheets, malformed .csv files, and unformatted .docx office memos:

• Merged headers and inconsistent date formats (2026/07/15, July 16, 2026, 17-07-26).
• Mixed currency strings ($1,250.00, EUR 450, 1250 USD).
• Duplicate vendor names (AMZN Mktp, Amazon Inc, amzn).
• Unstructured executive context locked inside Word documents.

Xberg Data Launderer pairs Xberg (the ultra-fast document intelligence engine) with local structured LLM extraction via Ollama (llama3.2:3b) to automatically parse, standardize, repair, and visualize office document bundles in under 10 seconds.

---

🛠️ Technology Stack

Layer	Tool / Library	Role
Document Parsing	Xberg	Extracts multi-sheet Excel tables, raw CSV data, and DOCX memo content
Local AI Engine	Ollama (llama3.2:3b)	Schema-driven JSON extraction, entity normalization, date standardization
Data Generation & Packing	Python zipfile, openpyxl, docx	Creates realistic messy office .zip bundles with embedded images
Output Visualizer	HTML5 / Modern Vanilla CSS	Standalone interactive dashboard with live search, filters, & statistics

---

🚀 Quick Start & Installation

1. Prerequisites

Ensure Python 3.10+ and Ollama are installed on your machine.

Download and start the lightweight local LLM model:

ollama run llama3.2:3b

2. Install Dependencies

Install xberg along with helper utility packages:

pip install xberg ollama openpyxl python-docx pandas pillow

3. Generate Sample Office Archive

Create the messy office ZIP package containing broken spreadsheets and a DOCX memo with embedded visuals:

python make_sample_zip.py

4. Run the Data Launderer Pipeline

Execute the end-to-end extraction, AI laundering, and dashboard generation:

python launder.py

5. View the Cleaned Dashboard

Open the generated HTML dashboard in your web browser:

Start-Process index.html

---

📂 File Structure

├── make_sample_zip.py       # Generates realistic messy .xlsx, .csv & .docx inside sample_office_data.zip
├── launder.py               # Clean Python script for extraction, AI laundering & dashboard rendering
├── dashboard_template.html  # Standalone HTML dashboard template with dark-mode styling & filters
├── sample_office_data.zip   # Sample office archive containing dirty spreadsheets & executive memo
├── cleaned_data.json        # Structured, validated JSON output produced by Ollama & Xberg
├── cleaned_summary.csv      # Exportable clean CSV summary table
└── index.html               # Interactive dark-mode HTML visualizer & executive dashboard

---

📊 Sample Workflow Pipeline

flowchart LR
    A[sample_office_data.zip] --> B[Xberg Engine]
    B -->|XLSX / CSV / DOCX| C[Raw Document Extraction]
    C --> D[Ollama llama3.2:3b]
    D -->|Structured Schema| E[Clean JSON / CSV]
    E --> F[index.html Visual Dashboard]

Loading

---

💼 5 Real-World Enterprise Use Cases

1. Quarterly Expense Audit: Clean inconsistent credit card statements from multiple employee CSV files without writing custom regex rules.
2. Supplier Invoice Reconciliation: Standardize vendor naming variants (Amazon, AMZN Mktp, Amazon Web Services) across thousands of invoices.
3. M&A Due Diligence Data Intake: Parse mixed Zip archives containing financial models and executive memos into structured analytics schemas.
4. CRM Contact Ingestion: Standardize customer names, phone numbers, and transaction amounts gathered from disparate regional sales spreadsheets.
5. Automated Executive Reporting: Synthesize numerical tables and Word memo summaries into a unified executive dashboard automatically.

---

🔮 5 Future Roadmap Features

1. Multi-Currency FX Conversion: Real-time currency normalization using live exchange rate APIs.
2. Automated Anomaly & Fraud Detection: Flag unusual transaction amounts or suspicious duplicate invoices using AI heuristics.
3. Multi-Model Fallback Chains: Seamless fallback from local Ollama to remote cloud providers for massive batch workloads.
4. Direct ERP Integration: Push cleaned structured records directly into SAP, QuickBooks, or Salesforce APIs.
5. Interactive Data Editing Interface: Allow manual inline corrections in the HTML dashboard before final ERP sync.

---

🏷️ Keywords & Tags

Xberg Document Intelligence OCR Local AI Ollama llama3.2 Python Spreadsheet Extraction Data Launderer Automate Boring Office Job Data Sanitization Excel Automation CSV Repair Word DOCX Extraction

---

📜 License

Distributed under the MIT License.