Ollama Scripts

A collection of scripts leveraging Ollama models for automation and productivity.

Scripts

propose-labels.py

Automatically suggest and apply labels to GitHub issues using Ollama LLMs.

Features:

• 🤖 AI-powered label suggestions with confidence scores
• 🎯 Multi-model consensus for validation
• 💬 Analyzes issue title, body, and comments
• 🎨 Color-coded terminal output
• 🔍 Smart filtering (unlabeled, needs triage)
• ✨ Configurable confidence thresholds
• 🚀 Can automatically apply labels

Requirements:

• Python 3.11+
• uv (automatically installs dependencies)
• Access to an Ollama server

Quick Start:

# Make script executable
chmod +x propose-labels.py

# List available models
./propose-labels.py --list-models

# Analyze issues (dry-run, no changes)
./propose-labels.py owner/repo --limit 5

# Process unlabeled issues
./propose-labels.py owner/repo --unlabeled --limit 10

# Use multiple models for consensus
./propose-labels.py owner/repo --models "phi3.5:3.8b,llama3.1:8b"

# Apply labels (requires GitHub token with write access)
./propose-labels.py owner/repo --apply --github-token $GITHUB_TOKEN

Options:

usage: propose-labels.py [-h] [--github-token GITHUB_TOKEN] [--model MODEL]
                         [--models MODELS] [--no-comments]
                         [--ollama-url OLLAMA_URL] [--issue-number ISSUE_NUMBER]
                         [--limit LIMIT] [--unlabeled] [--needs-triage]
                         [--confidence CONFIDENCE] [--ignore-labels IGNORE_LABELS]
                         [--apply] [--existing-boost EXISTING_BOOST]
                         [--verbose] [--list-models]
                         repo

Propose GitHub issue labels using Ollama

positional arguments:
  repo                  GitHub repository in format owner/name

optional arguments:
  --github-token        GitHub personal access token
  --model               Ollama model name (default: phi3.5:3.8b)
  --models              Comma-separated list of models for consensus
  --no-comments         Exclude issue comments (included by default)
  --ollama-url          Ollama server URL (default: https://llm.sbr.pm)
  --issue-number        Process only a specific issue number
  --limit               Number of issues to process (default: 5)
  --unlabeled           Only process issues with no labels
  --needs-triage        Only process issues missing area/ or kind/ labels
  --confidence          Minimum confidence threshold 0-100 (default: 75)
  --ignore-labels       Patterns of labels to ignore (default: priority/*,size/*,etc.)
  --apply               Actually apply labels (requires --github-token)
  --existing-boost      Confidence boost for existing labels (default: 20)
  -v, --verbose         Show verbose output
  --list-models         List available models from Ollama server

Examples:

# Triage workflow: find unlabeled issues and suggest labels
./propose-labels.py tektoncd/pipeline --unlabeled -v

# High-confidence only
./propose-labels.py tektoncd/pipeline --confidence 90

# Multi-model validation before applying
./propose-labels.py myorg/myrepo \
  --models "phi3.5:3.8b,llama3.1:8b,qwen2.5-coder:7b" \
  --needs-triage \
  --apply \
  --github-token $GITHUB_TOKEN

Default Ignored Labels:

The script automatically ignores automation and process labels:

• priority/* - Priority decisions (human)
• size/* - Size labels (automation)
• approved, lgtm - PR approval
• needs-ok-to-test - PR automation
• release-note* - Release notes
• cla* - CLA labels
• lifecycle/* - Lifecycle management
• do-not-merge/* - Merge control

License

MIT License - see LICENSE file for details

Contributing

Contributions welcome! Feel free to open issues or pull requests.