type: decision
status: active
timestamp: 2026-07-03
confidence: high
durability: durable
tags: [okf, upstream, google, spec]
status: active
timestamp: 2026-07-03
confidence: high
durability: durable
tags: [okf, upstream, google, spec]
OKF v0.2 additions upstreamed to Google
PR to GoogleCloudPlatform/knowledge-catalog proposing optional `confidence` and `durability` fields; agentmemory precedent cited.
OKF v0.2 → upstream to Google
Decision
Open PR/issue on GoogleCloudPlatform/knowledge-catalog proposing OKF v0.2 additive fields:
confidence: high | medium | low— matches rohitg00/agentmemory precedentdurability: durable | volatile— novel; oriz first-mover
Both optional. v0.1 files stay conformant.
Why upstream (not fork or private)
- v0.1-only status confirmed by deep-research 2026-07-02 — no official v0.2, no competing forks, only ~3 weeks old.
- First-mover window open — spec author (Google Data Cloud team) accepting community input.
- Ecosystem alignment — private v0.2 causes drift when Google eventually ships v0.2 of their own.
- Real precedent for
confidence(agentmemory) — not speculative.
Outcome (2026-07-03)
Filed:
- Comment on #151 — adopter data point (828 files) + variant choice (3-level lowercase vs their 4-level). https://github.com/GoogleCloudPlatform/knowledge-catalog/issues/151#issuecomment-4872064932
- New issue #172 —
durability: durable | volatileproposal (novel; not covered by #151 lifecycle or #160 maturity).
Not filed: separate PR/issue for confidence — already covered by #151. Filing duplicate would burn maintainer goodwill per terse-issues-less-hallucination rule.
What we ship in the community
- Diff to
okf/SPEC.mdadding two field defs. - Example bundle using both fields (small — 5 files).
- Rationale: hallucination reduction, durability triage for compaction/pruning.
- Cite agentmemory + oriz workspace (792 concept files) as adopters.
Terse per rule
Body ≤150 words per terse-issues-less-hallucination. Thank maintainers per thank-maintainers. No speculation about “future v0.3” — just the 2 fields.
Fallback
If Google rejects or ignores:
- Keep v0.2 internally.
- Rename
format_version: okf-v0.2→okf-v0.2-orizto avoid future collision. - Continue with additive-only additions.
Anti-patterns
- ❌ Fork spec repo aggressively — burns goodwill.
- ❌ Ship 5+ fields — dilutes proposal, higher rejection risk.
- ❌ Skip the PR — locks us out of official evolution.