A 22-chapter, group-organized, hands-on playbook covering everything from "what is LinkML" to "how do I run an LLM RAG agent over the Cytognosis KG". This is the standard reference for any new schema, ontology mapping, or KG ingestion task at Cytognosis.
The playbook itself is a directory of markdown chapters and supporting code, not a single document. It lives in the workspace folder rather than this repo because it includes large data examples and is regenerated when chapters are restructured.
Infrastructure and Tooling/linkml_kg_playbook/ (workspace folder)This pointer document captures the structure and the decisions that came out of the playbook, so the operational docs in this repo (the scholarly KG schema, SSSOM stack, paper library) can reference specific chapters reliably.
| Chapter | Topic |
| --- | --- |
| 00 | Setup |
| 01 | LinkML in an Hour |
| 02 | Schema Landscape (BibTeX, OpenAlex, Biolink, GA4GH, etc.) |
| Chapter | Topic |
| --- | --- |
| 03 | schema.org + Bioschemas |
| 04 | Biolink + core LinkML schemas |
| 05 | UMLS + SNOMED CT — RF2 / RRF coverage; key components; OLS4 release; snomed-owl-toolkit vs. umls2rdf adapter comparison |
| 06 | BioThings + DDE |
| 07 | GA4GH (VRS, VA, Phenopackets, Pedigree, Beacon; cmungall/linkml-phenopackets) |
| 08 | SOSA / SSN |
| 09 | CELLxGENE |
| 10 | HDMF + NWB |
| Chapter | Topic |
| --- | --- |
| 11 | Open Targets |
| 12 | OpenAlex + BibTeX |
| Chapter | Topic |
| --- | --- |
| 13 | Hybrid-layout pattern; v0.5.0 refactor target for cytognosis_scholarly_kg.yaml |
| Chapter | Topic |
| --- | --- |
| 14 | Full SSSOM workflow with OLS4 SnomedCT (download, parse, validate, RDF, SNOMED → MONDO) |
| Chapter | Topic |
| --- | --- |
| 15 | BioCypher — architecture, ontologies, schema-config philosophy, adapters with Biolink / MONDO / SNOMED examples + Open Targets case study |
| 16 | Koza + Monarch ecosystem — Koza vs. Monarch distinction, monarch-ingest deep dive, Monarch KG dedicated subsection (about, build, schemas, hands-on download / Neo4j / OAK), phenopacket-ingest with Protobuf and LinkML transform variants |
| 17 | BioCypher vs. Monarch — side-by-side architecture, inputs (schema / ontology + identifier resolution), outputs (formats + automation), use cases (Open Targets — BioCypher; STRING / PPI — both with CROssBARv2 vs. string-ingest), Cytognosis decision matrix |
| Chapter | Topic |
| --- | --- |
| 18 | AnnData harmonization — 5-tier OAK → SSSOM → SapBERT → LLM → curator pipeline |
| 19 | LLM / RAG over the KG — biochatter, phenomics-assistant, graph-agent-devops, LangChain / LangGraph patterns (text-to-Cypher, agentic), end-to-end Cytognosis bio-agent example |
| 20 | Structured extraction — OntoGPT (LinkML-templated, OAK-grounded), Instructor (Pydantic + LLM), PydanticAI; hands-on LinkML → Pydantic → Instructor pipeline; spaCy / scispaCy hybrid pattern; tier comparison |
| Chapter | Topic |
| --- | --- |
| 21 | Appendix — quick-references, prefix tables, troubleshooting |
biochatter / phenomics-assistant for LLM access (chapter 19).
cytognosis_scholarly_kg.yaml._data/umls/ and _data/snomed/ (workspace folder). At the time the chapter was written those folders were not present; once they are, the chapter's hands-on commands run as-written.scholarly-knowledge-graph.md — the schema the playbook revolves around.sssom-cross-ontology-mapping.md — chapter 14 is the worked SSSOM example.paper-library-architecture.md — the sovereign library that feeds the OpenAlex + Zotero ingest in chapter 12.monday-resource-boards.md — the boards the playbook's ingest pipelines populate.Infrastructure and Tooling/linkml_kg_playbook/ (workspace folder)
Last Updated: May 2026
Status: 22 chapters in 8 groups; living document
Classification: Internal Use Only