Torino

Multi-agentic JIRA triage tool for Red Hat Satellite. Automates the Program Triage process using Claude AI agents (QE, Product Owner, Developer, Docs) that debate and reach consensus on issue classification.

Prerequisites

• Python 3.10+
• Claude Code CLI installed and authenticated
• A JIRA Cloud account with API token

Setup

1. Install Claude Code CLI

npm install -g @anthropic-ai/claude-code

Configure authentication for your environment (e.g. for Vertex AI):

export CLAUDE_CODE_USE_VERTEX=1
export CLOUD_ML_REGION=global
export ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project-id
gcloud auth application-default login

For best results, run claude, type /model and select Opus 4.6 (1M context).

Verify it works:

claude --version

2. Clone and install Torino

git clone https://github.com/lhellebr/torino.git
cd torino
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

3. Configure JIRA credentials

cp config.example.yaml config.yaml

Edit config.yaml:

jira:
  server: https://your-instance.atlassian.net
  email: your.email@example.com
  api_token: your-jira-api-token

To create a JIRA API token, go to https://id.atlassian.com/manage-profile/security/api-tokens

Usage

# Triage specific issues
torino triage SAT-12345 SAT-12346

# Find and triage all untriaged issues (default: 5 newest)
torino triage

# Quick mode — single-agent classification, faster but less thorough
torino triage --quick SAT-12345

# Show each agent's assessment during the debate
torino triage -v SAT-12345

# Skip confirmation prompt and apply changes immediately
torino triage --yes SAT-12345

# Fetch more untriaged issues
torino triage --limit 20

# Use a different JIRA project
torino triage --project MYPROJ

# Combine options
torino triage --quick --yes --limit 10

What it does

For each issue, Torino:

1. Validates fields — checks if summary, priority, severity, component, regression, and description sections are filled
2. Searches for duplicates — uses AI-extracted keywords to find similar issues in JIRA
3. Runs a multi-agent debate — four AI agents (QE, Product Owner, Developer, Docs) independently analyze the issue, then challenge each other's assessments
4. Produces a consensus — a moderator synthesizes the debate into a single recommendation with reasoning
5. Applies to JIRA (after confirmation) — sets severity, priority, component, regression, labels, Need Info From, and posts a triage comment

Running tests

source .venv/bin/activate
pytest