Tip: select any text on this page to bring up the Ask tooltip and question the Synedrix OS assistant about it.
What Docling is
Docling parses PDFs, Office documents, HTML, and Markdown into a unified structural representation, preserving tables, nested headings, and reading order, rather than flattening a document straight to plain text.
Its Table Structure Recognition reconstructs a table's real header/row association from layout, not just its cell text: the specific capability this lab exists to test, confirmed under LAB-002's stand-up correctly parsing a real 14×7 table.
What it is normally used for
How Docling is used generally, before Synedrix OS enters the picture.
- RAG ingestion pipelines that need real document structure, not just extracted text
- Table-heavy document processing: financial reports, specs, anything where header/row association is load-bearing
- Multi-format normalisation: one converter across PDF/DOCX/PPTX/HTML/Markdown
What it owns, and what it does not
Every platform in the stack has a non-overlapping responsibility. The boundary is as much a part of the design as the capability.
Owns
- Nothing in production: reachable only through Semantica's DoclingParser wrapper inside LAB-002's isolated environment, never called directly
Does not own
- The production chunking/ingestion pipeline itself
- → The existing Markdown → Qdrant chunker in synedrix-os-website/agent, unmodified by this lab
- Semantic retrieval and similarity search: the single vector store
- → Qdrant (Layer 7); Docling's lab collection is additive, never a replacement
Why Docling
The one genuinely testable claim in Semantica's bundle
Everything else Semanticaships (RDF triple store, self-implemented graph algorithms, OWL/SHACL ontology) was pre-judged out of scope as store duplication or too large a question for the lab's timebox. Structural parsing is the exception: a falsifiable, measurable question, not an architectural judgment call.
A real first signal, not yet a robust one
LAB-002's retrieval-quality harness found recall@10 on multi-hop queries went 0.25 → 0.75 with Docling's table-aware chunking, with table-heavy and simple-lookup categories staying tied, at n=4 per category, a real signal that still needs a larger, repeated-run version before it is a robust one.
Licence position
Docling questions, answered
Short, citable answers drawn from this page.
What role does Docling play in Synedrix OS?
Docling (Structural document parsing) is mapped to Lab component: no OS layer in Synedrix OS. The one concrete, testable claim in Semantica's whole capability bundle: table-aware parsing, measured against the existing chunker.
What does Docling own in Synedrix OS?
Docling owns: Nothing in production: reachable only through Semantica's DoclingParser wrapper inside LAB-002's isolated environment, never called directly.
What does Docling not own in Synedrix OS?
Docling does not own: The production chunking/ingestion pipeline itself (owned by The existing Markdown → Qdrant chunker in synedrix-os-website/agent, unmodified by this lab); Semantic retrieval and similarity search: the single vector store (owned by Qdrant (Layer 7); Docling's lab collection is additive, never a replacement).
The full reference page, with the operating model, deployment detail, and links to every canonical specification this platform touches, lives in the documentation repository, which is private while the platform is being built.