v0.3.0

Pipeline validation and Tracer Overhaul

What's Changed

✨ Features

• feat(pipelines): tokenizer resource nodes — chunk_size counts real tokens by @Neeeser in #108
• feat(ingestion): never-fail embedding token guard with ingestion warnings by @Neeeser in #109
• refactor(pipelines): move tokenizer selection onto chunker config by @Neeeser in #112
• feat(dev): use ParadeDB for local development by @Neeeser in #114
• feat(traces): preserve result traceability across pipeline nodes by @Neeeser in #113
• feat(traces): streamline retrieval diagnosis by @Neeeser in #115

🐛 Bug fixes

• fix(pipelines): validate embedding limits from provider metadata by @Neeeser in #93

Full Changelog: v0.2.1...v0.3.0