zemmammehdi/port_tarrif_calculatorbis

An AI-powered solution to automatically extract, structure, and apply port tariff rules from the official South African port tariff PDF ) # 📄 South African Port Tariff Rule Extractor & Calculator This project automates the extraction of tariff rules from the official South African port tariff document (2024–2025) and uses them to calculate dues based on a user's vessel query. --- ## 🧠 Approach The solution is designed to be fully automated and agent-driven, based on the following pipeline: ### 1. PDF Parsing and Section Detection To automate the processing of the official tariff document, we first parse the PDF using a document parser. The **most critical step** is accurately **segmenting the document by sections**, which allows us to isolate rules related to specific tariff types (light dues, port dues, towage dues, etc.). ### 2. Gemini-Based Rule Extraction Once sections are detected and extracted, we use **Gemini LLM** to convert the unstructured text into structured **JSON rule files**, one for each port and each tariff type. This enables us to represent all tariffs programmatically, without hardcoding logic into the codebase. ### 3. User Input Parsing A natural language query describing the vessel (e.g., size, draft, DWT, activity, port) is parsed using another Gemini model to extract structured parameters. ### 4. Intelligent Orchestration A reasoning agent reads both: - The structured user input - The extracted JSON rules It then directly applies the correct formulas and logic from the rules to compute the tariffs. --- ## ▶️ How to Run ### ✅ Install dependencies ```bash pip install -r requirements.txt ``` ### ▶️ Execute the pipeline ```bash python main.py ``` This will: 1. Parse the tariff PDF and extract all relevant sections. 2. Generate JSON rule files using Gemini. 3. Accept a vessel input and parse its parameters. 4. Calculate all applicable tariffs using the extracted rules. ## 🔧 Tech Stack - Python 3.10+ - PyMuPDF / PDFPlumber (for PDF parsing) - Gemini (Google Generative AI) - JSON-based rule storage - Modular Python functions for tariff logic --- Let me know if …