CapEx/OpEx Classifier Agent

A finance analyst AI agent that classifies Jira work items as Capital Expenditure (CapEx) or Operational Expenditure (OpEx) using Qwen 2.5 32B via Ollama.

How It Works

The agent runs in three phases:

1. Context Loading — Feeds company classification standards into the model.
2. Classification — Batches Jira items and classifies each against the standards, returning structured JSON with confidence scores.
3. Reasoning Report — Generates a human-readable audit report explaining each decision.

Requirements

• Python 3.10+
• Ollama with the qwen2.5:32b model pulled
• Dependencies: ollama, pandas

Setup

# Pull the model
ollama pull qwen2.5:32b

# Install dependencies
pip install ollama pandas

Usage

1. Place your company standards in work-categories.csv (see included example).
2. Place your Jira export in jira-items.csv.
3. Run the classifier:

python capex-opex.py

Output is written to the output/ directory:

• <timestamp>_classified_items.csv — Full classification results.
• <timestamp>_classification_reasoning.txt — Auditor-friendly reasoning report.

Input File Formats

work-categories.csv

Column	Description
Category	Classification category name
Classification	CapEx or OpEx
Jira Issue Types	Allowed issue types (comma-separated)
Description	Category description
Logic/Rule	Keywords that trigger this category

jira-items.csv

Column	Description
Issue Key	Jira issue key (e.g., PROJ-101)
Summary	Issue summary/title
Issue Type	Story, Bug, Task, Epic, etc.
Description	(Optional) Detailed description

Configuration

Edit the CONFIG dict in capex-opex.py to change model, file paths, batch size, or context window.

License

MIT