Summary
Offering an integration: graphify-okf-bridge is a standalone tool that exports graph.json to/from Google's Open Knowledge Format (vendor-neutral markdown+YAML knowledge bundles), plus a linker that infers cross-silo edges between code nodes and OKF data/table concepts (dbt ref()/source(), SQL literal matching, exact-name fallback — all INFERRED confidence, precision-first).
Concretely this lets a merged graph answer things like graphify path "stg_orders.sql" "customers", ending in a reads_from [INFERRED] edge that crosses from application/pipeline code into a BigQuery table concept imported from an OKF bundle (e.g. Google's own GA4 sample bundle). Demo'd end-to-end against a real dbt-labs/jaffle-shop checkout + the GA4 bundle — see demo/.
What it does today (standalone)
okf-bridge export <graph.json> <bundle-dir> — Graphify graph → OKF bundle, passes okf-bridge validate --strict.okf-bridge import <bundle-dir> <graph.json> — any OKF bundle (including foreign/official ones, e.g. GA4) → a graph.json mergeable via graphify merge-graphs.okf-bridge link --repo-root <repo> <graph.json> <bundle-dir> — infers code↔data edges, output is self-contained so it composes cleanly with merge-graphs (each link() run emits both endpoints of every edge it produces, avoiding cross-file id-reuse issues we hit with merge-graphs's per-file namespacing).- Ships a
skill/SKILL.md (okf-bridge install-skill, mirroring graphify install) and documents the MCP path (Graphify's own graphify.serve already covers it — no second server needed).
Ask
Would an okf-format exporter/importer be something Graphify core wants in-tree (e.g. graphify export okf / graphify import okf), or is a standalone bridge (current state, already on PyPI as graphify-okf-bridge) the preferred shape? Happy to contribute a PR either way — wanted to check appetite before proposing a specific integration surface, since this touches merge-graphs composability and might warrant design input from maintainers first.
Summary
Offering an integration: graphify-okf-bridge is a standalone tool that exports
graph.jsonto/from Google's Open Knowledge Format (vendor-neutral markdown+YAML knowledge bundles), plus a linker that infers cross-silo edges between code nodes and OKF data/table concepts (dbtref()/source(), SQL literal matching, exact-name fallback — allINFERREDconfidence, precision-first).Concretely this lets a merged graph answer things like
graphify path "stg_orders.sql" "customers", ending in areads_from [INFERRED]edge that crosses from application/pipeline code into a BigQuery table concept imported from an OKF bundle (e.g. Google's own GA4 sample bundle). Demo'd end-to-end against a realdbt-labs/jaffle-shopcheckout + the GA4 bundle — seedemo/.What it does today (standalone)
okf-bridge export <graph.json> <bundle-dir>— Graphify graph → OKF bundle, passesokf-bridge validate --strict.okf-bridge import <bundle-dir> <graph.json>— any OKF bundle (including foreign/official ones, e.g. GA4) → a graph.json mergeable viagraphify merge-graphs.okf-bridge link --repo-root <repo> <graph.json> <bundle-dir>— infers code↔data edges, output is self-contained so it composes cleanly withmerge-graphs(eachlink()run emits both endpoints of every edge it produces, avoiding cross-file id-reuse issues we hit withmerge-graphs's per-file namespacing).skill/SKILL.md(okf-bridge install-skill, mirroringgraphify install) and documents the MCP path (Graphify's owngraphify.servealready covers it — no second server needed).Ask
Would an
okf-format exporter/importer be something Graphify core wants in-tree (e.g.graphify export okf/graphify import okf), or is a standalone bridge (current state, already on PyPI asgraphify-okf-bridge) the preferred shape? Happy to contribute a PR either way — wanted to check appetite before proposing a specific integration surface, since this touchesmerge-graphscomposability and might warrant design input from maintainers first.