README: TxnResolver

Overview

This project automates the categorization and resolution of financial transactions using AI-powered classification with the OllamaLLM model. It processes transaction data, categorizes cases, determines resolution status, and executes a workflow using a graph-based approach.

Prerequisites

Ensure you have the following installed before running the script:

• Python (>=3.8)
• pip (>=21.0)

Installation

1. Clone the repository:

   git clone <repository_url>
   cd <repository_folder>

2. Install dependencies:

   pip install -r requirements.txt

   Ensure requirements.txt includes:

   pandas
   chardet
   langchain
   langgraph
   langchain_ollama
   

3. Create necessary directories:

   mkdir -p uploads/not_found_sys_b uploads/processed uploads/pending uploads/resolved

Setting Up Ollama and Mistral

Install Ollama

Ollama is required to run the LLM. Install it using:

curl -fsSL https://ollama.com/install.sh | sh

Download and Install Mistral Model

To use the Mistral model with Ollama, run:

ollama pull mistral

Verify Installation

To check if Ollama is running and the Mistral model is installed, execute:

ollama list

This should display mistral as an available model.

Configuration

Modify the script to specify the correct paths for input files:

• recon_data_reply.csv
• recon_data_raw.csv

These CSV files should contain transaction data with necessary columns such as Transaction ID, Comments, recon_status, etc.

Running the Script

Execute the script with:

python script.py

Expected Output

• Categorized transaction files saved in uploads/
• Processed logs stored in process.log
• Resolved and unresolved transactions moved to corresponding directories
• AI-generated recommendations for unresolved transactions

Logging

All execution logs are stored in process.log, which helps track errors and process flow.

Troubleshooting

• Ensure input CSV files exist and have the correct column headers.
• Check process.log for any error messages.
• Ensure OllamaLLM is correctly installed and running.
• Verify that the Mistral model is available by running ollama list.

Contributions

Feel free to submit issues or pull requests to enhance the project!

License

This project is licensed under the MIT License.