Release Notes for 0.88.0
✨ Features
- Added support for
file://
prefix for local file paths in thetests:
field in configuration (@mldangelo) - Added
--env-path
as an alias for the--env-file
option in CLI (@mldangelo) - Introduced
PROMPTFOO_LIGHTWEIGHT_RESULTS
environment variable to optimize result storage (@typpo) - Added
validatePythonPath
function and improved error handling for Python scripts (@mldangelo) - Displayed custom named scores in the Web UI (@khp)
- Improved support for structured outputs in the OpenAI provider (@mldangelo)
- Added OpenAI Assistant's token usage statistics (@albertlieyingadrian)
- Added pricing information for Azure OpenAI models (@mldangelo)
- Improved API URL formatting for Azure OpenAI provider (@mldangelo)
🐛 Bug Fixes
- Fixed prompt normalization when reading configurations (@mldangelo)
- Resolved Docker image issues by adding Python, ensuring the
next
output directory exists, and disabling telemetry (@mldangelo) - Improved message parsing for the Anthropic provider (@mldangelo)
- Fixed error in loading externally defined OpenAI function calls (@mldangelo)
- Corrected latency assertion error for zero milliseconds latency (@albertlieyingadrian)
📚 Documentation
- Added a new Red Team introduction and case studies to the documentation (@typpo)
- Updated model references and default LLM models in the documentation (@mldangelo)
- Fixed typos and broken image links in the documentation (@mldangelo, @typpo)
🔨 Refactoring
- Refactored Red Team commands and types to improve code organization (@mldangelo)
- Moved
evaluateOptions
initialization toevalCommand
(@mldangelo) - Centralized cost calculation logic in providers (@mldangelo)
New Contributors
Full Changelog: 0.87.1...0.88.0