TinyTroupe Releases

A major version update upgrading the project from Paddle 2.5.1 to Paddle 3.0.0-beta, adding the AudioTools toolkit for DAC, and implementing loss functions required for DAC training.

Mar 5, 2025

This update fixes a bug related to validating nested schema fields and replaces the Google AI SDK dependency.

Feb 21, 2025

A maintenance release focused on fixing schema validation issues specifically within nested fields.

Feb 20, 2025

The release adds support for Google Gemini models, completes Ollama integration, includes Markdown in results, and allows autoschema to accept natural language instructions.

Feb 18, 2025

OmniParser v2.0 introduces a larger and cleaner dataset, achieving 60% latency reduction and 39.6 average accuracy on ScreenSpot Pro, while supporting multiple large language models via OmniTool.

Feb 13, 2025

This version adopts semantic versioning and introduces new options such as PaddleOCR support, icon detection image size options, and torch inference mode.

Feb 13, 2025

Release v0.4.0-alpha expands persona specifications with deeper traits, introduces JSON file support for agents, fragments for reuse, logical propositions, and intervention mechanisms.

Jan 29, 2025

This update expands supported file formats to include DOC, TXT, PNG, JPG, and HTML, and allows schemas to include boolean and enum field types.

Jan 14, 2025

This release introduces support for TimesFM-2.0 models with hyperparameter tuning capabilities and fixes bugs in PyTorch decoding. The median head is now set as the default point forecaster.

Dec 31, 2024

Introduces the ability to auto-generate schemas and adds file converters/formatters to the project.

Dec 14, 2024

The release adds PyTorch support for TimesFM inference with separate dependencies for the `pax` and `torch` versions, accompanied by a PyTorch checkpoint for the 200m model.

Oct 18, 2024

This release introduces new Speech-to-Text models like WavLM and HuBERT, expands Text-to-Speech capabilities with DiffSinger and JETS, and adds XPU support for FastSpeech2. It also includes various bug fixes, performance improvements, and documentation updates.

Jun 27, 2024

Minimal release notes: "# Others - fix typeguard version. https://github.com/PaddlePaddle/PaddleSpeech/pull/3056 @yt605155624"

Apr 14, 2023

This release focuses on expanding Cantonese and English support with new models like VITS and StarGANv2-VC, while adding support for multiple inference backends (Static, ONNX, PaddleInference) and removing deprecated Fluid APIs.

Mar 15, 2023

This release introduces major Speech-to-Text advancements with Whisper and Wav2Vec2 models, along with streaming TTS capabilities and end-to-end prosody prediction for mixed-language synthesis.

Dec 14, 2022