Tpane

tpane

The reference implementation of the TOPAZ (Test Output Protocol for AI Zealots) standard

Why "tpane"?

The name has a dual meaning that perfectly captures the tool's purpose:

1. T-Pain Auto-tune Analogy Just like T-Pain's auto-tune transforms raw vocals into polished audio, tpane transforms verbose test outputs into clean, structured formats optimized for AI analysis.

2. Test Output for Context Window tpane creates a clear "window pane" into your test results - but more importantly, it optimizes them to fit within AI context windows. No more truncated logs or overwhelming verbosity when analyzing test failures.

Key Benefits

• 60-80% token reduction with TOPAZ v0.3 format (66% improvement over v0.2)
• Structured, predictable format for reliable AI parsing
• Language-agnostic design for cross-framework adoption
• Progressive disclosure based on context and budget constraints

Quick Start

# Install dependencies (requires Python 3.9+)
pip install pyyaml

# Convert JUnit XML to TOPAZ format
python src/tpane.py --format junit test-results.xml

# Process pytest output with summary mode
pytest | python src/tpane.py --format pytest --mode summary

# Auto-detect format and use failures mode (default)
python src/tpane.py test_output.txt

Usage Examples

Basic Usage

# Auto-detect format and show failures (default mode)
tpane test_output.txt

# Specify format explicitly
tpane --format junit test-results.xml
tpane --format pytest pytest_output.txt
tpane --format rspec rspec_results.json
tpane --format tap tap_output.txt

# Process from stdin
pytest | tpane --format pytest
npm test | tpane --format tap

Focus Modes

# Show summary only (minimal token usage)
tpane --mode summary test_output.txt

# Show critical failures only (errors and key failures)
tpane --mode critical test_output.txt

# Show all failures (default)
tpane --mode failures test_output.txt

# Show only the first failure (for quick debugging)
tpane --mode first-failure test_output.txt

Token Management

# Set custom token budget (default: 5000)
tpane --limit 1000 large_test_output.xml
tpane --limit 10000 comprehensive_results.json

# Use TOPAZ v0.3 format (default)
tpane --topaz-version v0.3 test_output.txt

# Use legacy v0.2 format
tpane --topaz-version v0.2 test_output.txt

# Handle large files (default max: 50MB)
tpane --max-input-size 100 very_large_results.xml

About TOPAZ

TOPAZ (Test Output Protocol for AI Zealots) is a standardized test output format designed specifically for LLM consumption. It addresses the token efficiency, structured parsing, and cross-tool integration needs of AI-powered development workflows.

TOPAZ Design Principles:

1. Token Efficiency: Minimize tokens while preserving semantic completeness
2. Structured Data: Consistent schema for reliable programmatic access
3. Semantic Clarity: Clear causality (what failed and why) with actionable context
4. Progressive Disclosure: Multiple detail levels based on available budget
5. Cross-Framework: Language and tool agnostic design

TOPAZ Output Format

The tpane tool outputs YAML-formatted TOPAZ data. TOPAZ v0.3 (default) includes:

• Execution Context: Environment details for debugging (command, runtime, VCS info)
• Compact Format: Single-line structures for maximum token efficiency
• Focus Modes: Progressive disclosure from summary (50 tokens) to comprehensive
• Cross-Language Normalization: Consistent field names across test frameworks
• Smart Defaults: Only shows non-standard configurations

Example Output (TOPAZ v0.3 - Default)

EXECUTION_CONTEXT:
  command: python src/tpane.py --format pytest tests/
  pid: 12345 | pwd: /home/user/project
  runtime: python 3.11.5 (linux-x64)
  package_manager: pip 23.0.1
  vcs: git main@a1b2c3d
  test_framework: tpane (isolated)
  protocol: TOPAZ v0.3 | focus: failures | limit: 5000
  files_under_test: 24

tests/user_validation.py:
  L42: test failed
    Test: validates email format
    Expected: valid email
    Got: invalid@

Legacy Output (TOPAZ v0.2)

summary:
  total_tests: 156
  passed_tests: 142
  failed_tests: 12
  error_tests: 2
  execution_time: "23.45s"

file_results:
  - file_path: "spec/user_validation_spec.rb"
    test_count: 8
    failed_count: 2
    test_results:
      - name: "validates email format"
        passed: false
        expected: "valid email"
        actual: "invalid@"

Installation

# Install from PyPI (coming soon)
pip install tpane

# Or install from source
git clone https://github.com/delano/tpane.git
cd tpane
pip install .

For development setup, see [DEVELOPMENT.md](DEVELOPMENT.md). For deployment and release information, see [DEPLOYMENT.md](DEPLOYMENT.md).

Supported Test Formats

| Format | Auto-Detection | Status | |--------|----------------|---------| | JUnit XML | ✅ ` topaz_output.yaml

• name: Analyze failures with AI

uses: your-ai-analysis-action@v1 with: testresults: topazoutput.yaml

### CI/CD Pipeline
```bash
#!/bin/bash
# Run tests and process with tpane
npm test > test_output.txt 2>&1
python src/tpane.py --format tap test_output.txt > topaz_results.yaml

# Send to AI analysis service
curl -X POST https://your-ai-service/analyze \
  -H "Content-Type: text/yaml" \
  --data-binary @topaz_results.yaml

Technical Implementation

tpane implements the TOPAZ standard through:

• Smart Format Detection: Automatic identification of input test format
• Modular Parsers: Extensible parser architecture for new test frameworks
• Token Budget Management: Intelligent truncation and prioritization
• Path Normalization: Consistent file path handling across platforms
• Error Recovery: Graceful handling of malformed or incomplete test output

Contributing

We welcome contributions to both the TOPAZ standard and the tpane reference implementation:

• Standard Evolution: Propose enhancements to the TOPAZ specification
• Parser Extensions: Add support for new test frameworks
• Performance Optimization: Improve token efficiency and processing speed
• Integration Examples: Share real-world usage patterns

License

MIT License - see [LICENSE](LICENSE) file for details.

---

tpane - Because your test output deserves the auto-tune treatment! 🎤✨

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: delano
• Source: delano/tpane
• License: MIT

Install and usage instructions live in the source repository linked above.