From 06bb63feaa117b779c53e1b48ae555cdfc9cf189 Mon Sep 17 00:00:00 2001 From: harishghasolia07 <100846446+harishghasolia07@users.noreply.github.com> Date: Sat, 22 Aug 2026 06:07:17 +0000 Subject: [PATCH] Fix roster-count mismatch and document the missing ops-registry setup step README stated 'a team of 12' inside a sentence about this repo's own registry; the shipped example is sized for the 8 role archetypes, and 12 was production's headcount at measurement time. Clarified which is which. Separately, OPS_REGISTRY_PATH points at config/ops-registry.json, which isn't in the repo -- only the .example.json is. A live run (pull/poll/serve/board) hits ops-registry degraded mode with no doc explaining why. Added the copy step where live commands are first introduced. Verified: without the file, opsRegistryDegradedReason() returns 'nobody is in the roster'; with it copied from the example, null. --- README.md | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 18129b5..b590392 100644 --- a/README.md +++ b/README.md @@ -16,8 +16,9 @@ npm run demo -- --twice # a redelivery costs zero tokens From real trace data across 49 production runs — 48 meetings and one channel log — totalling 711 items, measured 2026-08-17: **14.5 items per run, 62.6% applied automatically, 27.3% held for a -human, 8.2% skipped as a duplicate, 2.0% failed.** It runs for a team of 12, which is the roster size -in the routing registry that governs it. +human, 8.2% skipped as a duplicate, 2.0% failed.** That production system's roster was 12 people at +measurement time. The registry shipped here (`config/ops-registry.example.json`) is sized for the +eight role archetypes below, not a headcount claim about this repo. Those four dispositions partition every item — 445 + 194 + 58 + 14 = 711, so a fifth outcome would show up as a gap. (The rounded percentages sum to 100.1; the counts are the claim.) @@ -189,9 +190,12 @@ termination, process supervision, queue durability, horizontal scale and OAuth t here, and every team's are different. **Commands join the sources seam to the pipeline**, so "this repo reads GitHub" is something you can -run rather than something you read: +run rather than something you read. Past the demo, these need a routing registry — the demo scenarios +each carry their own, but a live run reads `OPS_REGISTRY_PATH` (`.env.example`), which points at a +file that isn't in the repo: ```bash +cp config/ops-registry.example.json config/ops-registry.json # edit in your own roles, then npm run pull -- --source github --repo owner/name --since 2026-08-01 npm run pull -- --source gmail --thread npm run pull -- --source drive --file --write # --write, or it only plans