Hire your 24/7 engineer.(no sleep, no coffee)Pay him in tokens.(he might be dumb, but he's quite cute)
Dori lives in your chat. Tell him what to do in Slack, Discord or Telegram.He gives each job its own coding-agent session and reports back when it's merged.That's the whole setup.
No credit card. Early access and updates only.
5xx on POST /exports since 13:58. p95 latency 9.4s. Region: all.
M0 replies View thread
On it. @mina you're on-call, tagging you now. Investigation so far:
ok. check the logs now and ship a hotfix
Opened a ticket and took it.
DoriFound it in the logs: OOM on the buffer at 13:58:41, then the retry storm doubled the load.
PR #418 open: streams to object storage, drops the retry on timeout. Tests green.
QA on staging passed, a 3 GB export finished in 41s.
Hotfix deployed to all regions.
Error rate back to 0.3%, p95 1.2s. Fixed the cause, not just the symptom: no more full-workspace buffer.
Dori
Trusted by professionals at
Vercel
Cloudflare
TeamViewerWhat he does when you're not looking.
Works around the clock
He doesn't wait for a ping. He finds the next task himself.
Syncs the tickets
Notion and the ticket board stay current without you touching them.
Updates every hour
Lots of short updates. One living status message per job, edited in place.
Runs several sessions at once
Parallel jobs, one session each, every one tracked until it's merged.
The repo he maintains.
code-yeongyu/oh-my-openagent, Jul 13 to Oct 11, 2026. Counts are the public GitHub totals for the repo, all contributors.
Things people actually ask him.
Went through the voice note from the 14:30 sync. @mina @jun this touches your parts.
Follow-ups I can take: the export retry policy and the new rate-limit headers.
One knock-on: changing the retry policy means touching the queue worker we're mid-way through refactoring.
Should I go ahead now, or wait until the refactor lands?
@dori left comments on the session-store design doc, mostly about the eviction rules
On it, already reading!
1 reply Today at 10:34Applied all six. Eviction is now LRU with a 24h floor, and the two open questions have answers inline.
Research on queue backends is done. Four options compared on throughput, cost at our volume, and ops load.
ok this is actually really good
wait the charts lmao
@jun look at p.6 before the call
Working on the export limits and need a decision:
HHT-57 · Default export size cap- A. 2 GB cap, hard fail above
- B. 5 GB cap, stream above 2 GB (my pick: covers 99.4% of runs, no new infra)
- C. No cap, always stream
Other agents say 'done.'They lie.He pings only after the merge.
Watches your logs 24/7.Spots the regression.Hotfix shipped.
He spawns three Codex sessions.Codes through them. Ships their PRs.
Talk to Dori where you already are.
He messages as you, in the channel's language, the way you write. Short, no emojis.
Platform list as published on omo.dev. Dori mode today ships with Telegram, Discord and Slack.
waitlist hereand get early access
no credit card required.