open-bridge · live demo

What a real session looks like

Four flows from the instance that builds open-bridge — BKS-Lab's own. Branding and work are real; client names are kept generic. The formats are exactly what you'd see: the morning greeting, the briefing dashboard, an incident handed to a specialist, and first-run onboarding.

bks · the-bridge

① Morning — session start

$ cd ~/Developer/bks/the-bridge
⚡ Good morning, Alex. 🔥 5d 🔔 2
▸ BKS-Lab · the-bridge · user/alex
the-bridge user/alex ↑3 ±6
⟳ docs(open-bridge): launch strategy + live demo (1h ago)
🔴 2 waiting for your review ⚠ WIP 5/5
┌─ Doing — BKS-Lab ─────────────────────────────┐
│ ● open-bridge-launch distribution + live demo │
│ ● bridge-deck calendar timeline view │
│ ● bks-news weekly briefing automation │
│ +2 more │
└───────────────────────────────────────────────┘
$ claude
<bridge-phase0>
state=NORMAL · branch=user/alex · config=yes
3 CORE commits ahead — offer `git merge development`
</bridge-phase0>
Morning. Week 26: 5 in Doing — open-bridge launch leads, plus
bridge-deck and bks-news. 2 PRs wait on your review, WIP 5/5 at cap.
development is 3 commits ahead — merge when convenient. What's first?

② /briefing — the daily dashboard

> /briefing
╭──────────────────────────────────────────────────────────╮
│ Week 26 · Tue 24 Jun · 09:02 │ Doing: 5 │ ⚠ 2 │
│ Focus: ship the open-bridge launch demo → open the repo │
╰──────────────────────────────────────────────────────────╯
── ⚠ Warnings ──────────────────────────────────────────────
🟡 WIP 5/5 — at cap; close or reclassify before new work
🟡 Tailscale node-key expires in 18 days
── Focus today ─────────────────────────────────────────────
💻 open-bridge-launch distribution + live demo 🟢 doing
🔧 bridge-deck calendar timeline view 🟡 doing
📝 bks-news weekly briefing automation 🟢 doing
── GitHub (BKS-Lab) ────────────────────────────────────────
Agents & MCPs #20 4 open · 1 in review
BKS Operations #25 3 open
── Open PRs (2 human · 1 bot) ──────────────────────────────
#61 calendar timeline bridge-deck ↑ sam 1d
#18 launch demo open-bridge ↑ alex 3h
── Activity (7 days) ───────────────────────────────────────
open-bridge ▃▅▄▇▆█▅ 24 commits
bridge-deck ▂▃▅▄▆▃▄ 13 commits
bks-news ▁▂▁▃▂▄▂ 7 commits
── Calendar today ──────────────────────────────────────────
◆ 11:00 Weekly · BKS-Lab · 45m
── Next steps ──────────────────────────────────────────────
[ ] merge development (3 commits ahead)
[ ] review PR #61 (bridge-deck) → then it can land
[ ] renew the Tailscale node-key before it expires

③ Operator — find the bug, TDD the fix, ship

> inbound invoices aren't booking — what happened?
Loading the Peppol ops playbook. Dispatching the log-analyst against
the last 3h of fn-peppol-in-prd logs — read-only, isolated.
◐ Task · subagent_type=log-analyst · running …
◇ log-analyst → returned
Summary NEW BLOCKER — inbound parse fails in prod since 08:31 UTC; docs stuck in retry
Scope last 3h · fn-peppol-in-prd · 1,120 hits
Clusters 🔴 parse_error @ map_invoice 54× 08:31→now NEW
Correlation doc_8842 — received OK · parse FAIL @ map_invoice · never booked
Root cause payment_means code "59" (SEPA DD) missing from the enum → KeyError aborts the pipeline
That's a code bug in invoice-package, not the operator. Fixing it test-first.
▸ red — a test that reproduces doc_8842
+ tests/test_payment_means.py::test_sepa_dd_59
▸ pytest -k payment_means
✗ 1 failed — KeyError: '59' ← reproduced
▸ green — minimal fix
~ invoice-package/mapping.py · payment_means via .get(code, OTHER)
▸ pytest · ruff · mypy
✓ 214 passed · ruff ✓ · mypy ✓ · coverage 100%
▸ ship
✓ commit a3f1c9 · branch fix/payment-means-59 · pushed · PR #87
✓ deploy uat → fn-peppol-in-uat · health ok
✓ replayed doc_8842 on UAT → booked
Done: off-spec SEPA-DD code "59" hit a non-tolerant enum → KeyError, so every
inbound aborted. Test-first fix red→green, full suite + lint + types green,
shipped to UAT and replayed doc_8842 — it books cleanly now.
Prod deploy + PR #87 merge are your call. Ship it? [y/n]

④ First run — onboarding

> hi
Hi — I'm ELOPe.
I'm the AI running in The Bridge — your command hub for agents, repos
and standing orders. I watch your stack, dispatch specialists, help
you ship, and remember every session so tomorrow-you picks up here.
Looks like you're running me for the first time — fresh clone, nothing
wired up yet. That's a 5-minute fix.
Want me to onboard you? Guided and reversible. I'll:
1. spin up your user/alex branch (keeps your data out of CORE)
2. scan your projects folder and map the BKS-Lab ecosystem
3. pick a theme and wire up your first agents
[y] Let's go [t] Tell me more [n] Not now
> y
Great — about 5 minutes. [A] Identity
Name → detected "Alex" from git config. Keep it? [y/edit]
GitHub → bks-lab · Projects → ~/Developer/bks
✓ created + checked out branch user/alex
[B] May I look at your setup? (names + structure only, never content)
✓ developer repos across bks-lab + reinvent-lab
✓ homebrew gh, az, rclone, restic, tailscale, pandoc
✓ azure-cli logged in · 1 subscription
✓ tailscale 4 devices: mac-mini, macbook-pro, +2
[C] Based on what I found:
① rclone + restic detected → enable backup topology? [y/m/l]
✓ infra/backups/topology.yaml scaffolded · backup skill on
② tailscale (4 devices) → scaffold infra/remotes/? [y/m/l]

Live BKS-Lab instance — the team behind open-bridge. Branding, repos and stack are real; client names are kept generic and there is no customer data, no secrets, no PII. Formats mirror the live briefing, session-start, coordinator/sub-agent and bridge-onboard flows.