Repository navigation
[outreach] dibs.hivecommons.dev (MCP Registry remote) serves 10-19 build f55e314, 65 commits behind main #558
Description
Outreach Opportunity
Type: ecosystem-partnership (official MCP Registry listing)
Target: MCP clients that find Dibs through the official MCP Registry (dev.hivecommons/dibs) or the README's agent setup sections
The official MCP Registry lists dev.hivecommons/dibs as active (since 2026-10-19), and its only remote is https://dibs.hivecommons.dev/mcp. Anyone who installs Dibs from the registry or from the README gets whatever build is running there. That build is now well behind main:
| commit | date | |
|---|---|---|
Live (GET https://dibs.hivecommons.dev/healthz → version, checked 2026-11-02) |
f55e314c8662cbeb1056e01a76602c5e46f2df86 |
2026-10-19 |
main HEAD (all check-runs green; ghcr.io/hivecommons/dibs:7d2b67f9db62927723619d7acb262ac20d50cc4c published 2026-11-01) |
7d2b67f9db62927723619d7acb262ac20d50cc4c |
2026-11-01 |
git rev-list --count f55e314..7d2b67f = 65 commits. Users of the hosted endpoint don't yet have, among others:
- idea intake: RTF uploads keep paragraph breaks and decode
\uNescapes (🐛 fix(intake): skip RTF header destinations and keep paragraph breaks #425, 🐛 fix(intake): decode RTF \uN escapes and Windows-1252 bytes in StripRTF #544); HTML uploads no longer leak script/style/template text into the idea (fix(intake): drop style, script, template and comment text from HTML #476, fix(intake): parse HTML with golang.org/x/net/html so quoted attributes cannot leak #484, 🐛 fix(intake): drop script/style/noscript/template content after a self-closing opener #486); malformed multipart uploads return 400 instead of 413 (fix(intake): report malformed multipart uploads as 400, not 413 #430) - matching: zero-score CNCF candidates are no longer surfaced (🐛 fix(match): drop zero-score CNCF candidates before LLM scoring and surfacing #517), and legacy CNCF suggestion caches are invalidated (fix(match): invalidate legacy CNCF suggestion caches #521)
- MCP tools: idea URLs are derived from the request host (🐛 fix(mcpserver): derive idea URLs from the request host when X-Forwarded-Host is absent #539, 🐛 fix(mcpserver): derive the idea URL scheme from the request and sanitise forwarded headers #542)
- GitHub rate-limit back-off fixes across registry, news and history refresh (🐛 fix(registry): enrich descriptions once and back off on GitHub rate limits #480, 🐛 fix(history,news,catalog): honor GitHub rate-limit reset before polling again #488, 🐛 fix(ghrate,registry): prefer Retry-After over X-RateLimit-Reset when backing off #490, 🐛 fix(ghrate): cap every rate-limit back-off at MaxBackoff #551, 🐛 fix(ghrate): fall back to Cooldown when a rate-limit deadline is already past #554, 🐛 fix(registry): back off only on real GitHub rate limits, mark other 403s as looked #556)
Why the gap came back
#417 closed on 2026-10-19 after the Deployment was moved from :latest to a pinned sha (kubectl -n dibs set image ... dibs:<sha>), which deploy/README.md recommends ("Pin the Deployment to a sha for production"). A pinned Deployment never moves on its own, and nothing in the repo rolls it forward, so production stays at whichever sha was last set by hand. deploy/deployment.yaml in the repo still says :latest, so the manifest no longer matches what the cluster runs.
Proposed Action
Needs someone with access to the OKE cluster. The hive has no cluster credentials, so it can't do any of this.
- Roll production to the current green
mainimage, the same way [outreach] dibs.hivecommons.dev serves a 2026-09-09 build; roll out main before MCP Registry publication (#375) #417 was resolved:
kubectl -n dibs set image deploy/dibs dibs=ghcr.io/hivecommons/dibs:7d2b67f9db62927723619d7acb262ac20d50cc4c && kubectl -n dibs rollout status deploy/dibs --timeout=120s,
then confirmcurl -fsS https://dibs.hivecommons.dev/healthzreports7d2b67f9…and/readyzreturns 200. Before rolling, checkrunbooks/release-rollback.md§2 (store-format changes): none of the 65 commits change the idea/index file layout as far as their titles show, but the operator should confirm.
Recommendation (owner choice): pick a rollout cadence for the pinned sha, e.g. after each batch of fix(intake|match|mcpserver) merges or weekly. Without one, this issue will keep coming back (#417 was the same gap at 175 commits).
Risk: infra (production rollout on a cluster only the owner can reach; public endpoint behind the MCP Registry listing)
Filed by outreach agent (ACMM L6 — full mode)
— hive: agent=outreach backend=copilot model=claude-opus-5.5 copilot=1.0.88
Activity
hivecommons-hive commentedon Nov 2, 2026
What to reply
/hive approve: go with the recommendation/hive decision <your instructions>: go ahead, following your instructions/hive help: show these againIf what this issue waited on has been done (a merged fix, or an action taken outside this repository), a maintainer can reply
/fixedto close it as completed; Hive clears its park labels.Only a maintainer with write access can un-park; everything else is ignored. Full command reference: https://github.com/hivecommons/hive/blob/v5/src/docs/maintainer-commands.md.