Kreuzberg v4.7.0 boosts code intelligence

This is the kind of release that moves a parser from "useful" to "pipeline-critical": Kreuzberg is clearly optimizing for agent workflows, not just clean text extraction.

• –AST-aware code chunking and symbol extraction make it more viable for code indexing, PR review, repo search, and MCP-driven agents
• –The benchmark story matters: structural correctness is the real product here, and the reported jumps in LaTeX, XLSX, and PDF tables suggest serious extraction work rather than cosmetic polishing
• –Unified typed documents plus multiple renderers reduce format drift, which is exactly what downstream LLM systems need if they’re going to trust the output
• –OpenWebUI integration broadens adoption beyond library users and positions Kreuzberg as infrastructure for self-hosted AI stacks
• –TOON output is a pragmatic token-saving move, but the bigger win is that the release treats output shape, validation, and security as first-class concerns