Summary
Google Cloud recently published the Open Knowledge Format (OKF): an open, vendor-neutral spec for representing curated knowledge as markdown files with YAML frontmatter, linked into a graph via standard markdown links. No new SDK, no new runtime — just files.
(ref: https://cloud.google.com/blog/products/data-analytics/how-the-open-knowledge-format-can-improve-data-sharing)
I'd like to propose an OKF Adapter Skill for agno-dev-toolkit: a lightweight, read-only tool that lets any Agno agent load an OKF bundle as structured, typed context.
Motivation
- OKF sits in the gap between raw vector RAG (good for unstructured text) and a full GraphRAG setup (heavy to run) — it's a git-friendly, versionable knowledge layer that fits how this toolkit already treats validation rules (
ValidationEngine + ValidationRule ABC).
- Zero vendor lock-in, consistent with the toolkit's existing stance (ChromaDB default, Milvus opt-in via
VECTOR_BACKEND, stdio MCP transport).
- Being an early, genuinely useful OKF↔Agno integration is good for the project and the wider
built-with-agno community — not a marketing move, an actual gap this fills.
Proposed scope (MVP)
Out of scope (v1)
- Writing/updating OKF bundles from within an agent
- Vector search over OKF content (that's what ChromaDB is for)
- Any new server process — stays a pure parsing/adapter skill
Technical notes
- Fits the existing modular skill pattern; no changes to
start.sh required
- Optional
OKF_BUNDLE_PATH env var, following the toolkit's existing opt-in convention
Call to action
This is a proposal, not a commitment — feedback welcome before I start building:
- Does this belong as a core skill, or as a separate optional package?
- Anyone want to co-build or review a first draft PR?
- 👍 or comment if this is useful to you, and I'll open a draft PR with the MVP scope above.
Suggested labels: enhancement, discussion, help wanted
Summary
Google Cloud recently published the Open Knowledge Format (OKF): an open, vendor-neutral spec for representing curated knowledge as markdown files with YAML frontmatter, linked into a graph via standard markdown links. No new SDK, no new runtime — just files.
(ref: https://cloud.google.com/blog/products/data-analytics/how-the-open-knowledge-format-can-improve-data-sharing)
I'd like to propose an OKF Adapter Skill for
agno-dev-toolkit: a lightweight, read-only tool that lets any Agno agent load an OKF bundle as structured, typed context.Motivation
ValidationEngine+ValidationRuleABC).VECTOR_BACKEND, stdio MCP transport).built-with-agnocommunity — not a marketing move, an actual gap this fills.Proposed scope (MVP)
.mdfiles + YAML frontmatter) into typed concept objects (Pydantic)query_okf_concept(id),list_okf_concepts(tag)Out of scope (v1)
Technical notes
start.shrequiredOKF_BUNDLE_PATHenv var, following the toolkit's existing opt-in conventionCall to action
This is a proposal, not a commitment — feedback welcome before I start building:
Suggested labels:
enhancement,discussion,help wanted