What's Changed
Added
- LLM usage tracking — new
llm_usagefield onExtractionResultcaptures token counts, estimated cost (USD), model identifier, and finish reason for every LLM call (VLM OCR, structured extraction, LLM embeddings). Exposed across all 12 bindings.
Fixed
- Markdown chunker heading duplication when
prepend_heading_contextis enabled (#701) - Helm chart icon 404 on Artifact Hub —
.png→.svg - Python wheel manylinux compliance — bumped to
manylinux_2_39 - FFI memory leaks —
djot_content_json,structured_output_json,llm_usage_jsonnot freed - R e2e embed tests — missing
typediscriminator in generated config - Elixir parity test —
ExtractionConfigmissinghtml_outputfield - Go LLM e2e tests —
EmbeddingModelTypemissing LLM config support - WASM tree-sitter build — removed stale
wasmfeature gate for tslp 1.6.0 - Ruby binding compilation — magnus type inference errors and missing
llm_usagefield
Full Changelog: v4.8.4...v4.8.5