Changelog

All notable changes to Polysome will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

0.1.1 - 2025-12-23

Renamed Package: Canonical name changed from Polysome to polysome (lowercase) for consistency with Python standards.
GUI Integration: Prompt Editor is now integrated into the CLI via polysome-gui.
Dependency Structure: Introduced [ui] extra for minimal GUI install and [all] for full engine support.
Default Model: Updated example templates to use google/gemma-3-4b-it.

Jinja2 Syntax: Corrected variable interpolation syntax in prompt templates.
Workflow Error Handling: Improved Workflow.run to correctly report node failures.
CLI Scripting: Fixed Bash arithmetic evaluation in integration test scripts.

Initial Release: Polysome v0.1.0 is now available!
Workflow Engine: DAG-based pipeline orchestration for data processing and generation.
Inference Engines: Support for multiple LLM backends:
- HuggingFace Transformers: Standard backend for PyTorch models.
- vLLM: High-performance inference engine with quantization support.
- llama.cpp: Efficient CPU/Apple Silicon inference using GGUF models.
- vLLM Data Parallel: Multi-GPU batch processing support.
Node Types:
- load: Flexible data loading from JSON/JSONL/CSV.
- text_prompt: Core LLM processing node with Jinja2 templating and few-shot support.
- combine_intermediate_outputs: Merge workflow branches.
- Utility nodes: regex_split, sentence_split, deduplication, row_concatenation, column_concatenation.
CLI Tool: polysome command-line interface for:
- init: Scaffolding new projects with templates.
- run: Executing workflows with validation.
Docker Support: Ready-to-use Docker images for GPU (CUDA) and CPU (ARM64) environments.
Prompt Editor: Streamlit-based UI for designing and testing prompts (polysome-prompt-editor).
Documentation: Comprehensive guides for installation, workflows, prompt engineering, and docker usage.