Kiln-ai

⬇️ Download links in the "Assets" section below ⬇️

⬇️ Download links in the "Assets" section below ⬇️

This release adds a ton of new integrations including 5 new AI providers, 5 new inference models, 3 new fine-tune base-models

- New AI Providers: **Gemini API, Vertex AI, Hugging Face, Azure OpenAI and Together.ai**
- New Inference Models: **QwQ, Gemma 3, o1, o3, Phi 4 Mini**
- **New fine-tuning host: Together.ai** which adds new serverless fine tuning models: Qwen 2.5 14B, Qwen 2.5 72B, Llama 3.2 1B
- Features:
- **Import existing datasets via CSV** (thanks leonardmq)
- **Weights and Biases integration** for fine tuning metrics
- Allow editing titles/descriptions throughout the UI
- Add UI for deleting evals, tasks, custom prompts, and other data model items directly from the UI
- Core tech
- Migrate to LiteLLM, removing the last of LangChain
- Python 3.13 support (and ready for 3.14)

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/13956877583
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/13957266813

**Full Changelog**: https://github.com/Kiln-AI/Kiln/compare/v0.12.1...v0.13.0

⬇️ Download links in the "Assets" section below ⬇️

New in this release **Evals - a comprehensive evaluation toolkit**:
- Build powerful SOTA evals (G-Eval, LLM as Judge)
- Check eval correlation to human preferences (Kiln Ratings) to find the best evaluator
- Synthetically generate eval datasets with Kiln Synthetic Data Gen
- Use the analysis tools to find the optimal prompt+model for your task
- Automatic eval: Kiln will automatically build an eval for any Kiln task using your task definition
- Templates for common eval use cases: bias, toxicity, jailbreaking, maliciousness, factual correctness.

Other new features include:
- Support for distilling (fine-tuning) an open model from Sonnet 3.7 thinking
- New Built-In Models: Sonnet 3.7, Dolphin 2.9 8x22B, and Grok
- Improved logging (thanks to leonardmq)
- ARM Linux builds now included

CI Build Source for this release Mac and Linux: /Kiln-AI/Kiln/actions/runs/13638728833
CI Build Source for this release Windows: /Kiln-AI/Kiln/actions/runs/13638728765

**Full Changelog**: https://github.com/Kiln-AI/Kiln/compare/v0.11.1...v0.12.1

⬇️ Download links in the "Assets" section below ⬇️

This release focuses on making Kiln easier to use, across lots of small fixes.

**New features include:**

- **Synthetic data generation is 25x faster**
- **Windows Code Signing** which means fewer "Unrecognized app" prompts, and virus false positives
- **Resolve a bugs causing a blank screen** on some Windows 11 machines
- **Improved Ollama support** when using structured data (JSON)
- You can disconnect AI providers from settings UI, without editing config files
- Improved docs, and links to the docs from within the app
- Add the ability to clone a task, copying all it's fields
- Show synthetic dataset topic in the datasets view

CI Build Source for this release Mac and Linux: Kiln-AI/Kiln/actions/runs/13234579948 and Windows: Kiln-AI/Kiln/actions/runs/13234579954

**Full Changelog**: https://github.com/Kiln-AI/Kiln/compare/v0.10.1...v0.11.1

⬇️ Download links in the "Assets" section below ⬇️

Note: v0.10.1 is a minor release with a bug fix that blocked new users from running models. The remainder of the release notes are features new in v0.10 which shipped yesterday.

**New features include:**

- **Add support for reasoning models, including the ability to distill your own reasoning model.** [Docs](https://docs.getkiln.ai/docs/reasoning-thinking-and-chain-of-thought)
- Many new models: R1, R1 distills, Mistral Small 3, Phi 4, and Gemini 2.0 flash
- Improved structured data generation (JSON)
- Fine-tuning prompt alignment: use the prompt system the model was tuned with by default.
- Support for the new faster Fireworks V2 fine-tuning system

CI Build Source for this release: Kiln-AI/Kiln/actions/runs/13192037218

**Full Changelog**: https://github.com/Kiln-AI/Kiln/compare/v0.10.0...v0.10.1

⬇️ Download links in the "Assets" section below ⬇️

**New features include:**

- **Add support for reasoning models, including the ability to distill your own reasoning model.** [Docs](https://docs.getkiln.ai/docs/reasoning-thinking-and-chain-of-thought)
- Many new models: R1, R1 distills, Mistral Small 3, Phi 4
- Improved structured data generation (JSON)
- Fine-tuning prompt alignment: use the prompt system the model was tuned with by default.
- Support for the new faster Fireworks V2 fine-tuning system

CI Build Source for this release: Kiln-AI/Kiln/actions/runs/13148150574

**New Contributors**
* SimonCoulombe made their first contribution in https://github.com/Kiln-AI/Kiln/pull/109
* bvqbao made their first contribution in https://github.com/Kiln-AI/Kiln/pull/117

**Full Changelog**: https://github.com/Kiln-AI/Kiln/compare/v0.9.0-beta...v0.10.0-beta