Search Tools, Documents & RAG
We're excited to launch our biggest project ever: Kiln Search Tools. It's an easy-to-use but powerful tool for building state-of-the-art RAG systems (Retrieval-Augmented Generation). It's as easy as dragging and dropping documents into Kiln. It only takes a few minutes to enhance your AI task's performance with custom search.
Highlights include:
- Easy to get started: just drop in documents, select a template configuration, and you're up and running in a few minutes. We offer several one-click templates for state-of-the-art RAG pipelines.
- Highly customizable: advanced users can customize all aspects of the RAG pipeline to find the ideal RAG system for their data. This includes the document extractor, chunking strategy, embedding model/dimension, and search index (vector/full-text/hybrid).
- Wide Filetype Support: Search across PDFs, images, videos, audio, HTML and more using multi-modal document extraction.
- Document Library: manage documents, tag document sets, preview extractions, sync across your team — all from an easy-to-use UI.
- Team Collaboration: Documents can be shared with your team via Kiln's Git-based collaboration.
- Deep Kiln Integrations: works with Kiln's evals, datasets, AI providers, tool calling, and more.
Check out the search tool documentation for everything you need to get started.
This is just the start! RAG is a complex problem with many design choices. Kiln will keep adding more options, keeping it easy for those new to AI, while giving powerful options to those who want to go deep on optimization. If you have feature requests, please let us know on our Discord.
Synthetic Data Generation V3
We also just launched a new update to Synthetic Data Generation! The V3 update makes several improvements to our data gen system:
- New step-by-step guidance on the synthetic data generation process
- More curation options: preview model outputs before saving
Check out the synthetic data generation documentation for details.
Kiln v0.21.1 App Release
Our new v0.21.1 app release includes:
- Search Tools & RAG
- Synthetic Data Generation V3
- New 'Trace' UI visualizing all steps of AI calls, including tool-calls/responses, and internal thinking steps
- New Models: Qwen 3 Max, Qwen 3 Next, Kimi K2 0905, Seed OSS 36B, Step3
- And More: UI improvements, embedding models, LanceDB integration, llama_index integration, and more
