The Brain↔Pulse A2A bridge — measured, and one defect in your config

Young, 2026-10-12: "Might need to enable A2A to establish bridge between Pulse and Brain Unit."

Measured first. The bridge already exists and works — I used it live. This is not an enablement task. It is one URL defect in your profile, plus a gap in mine and Sol's.

1. What already works — proven, not assumed

The serving LAB gateway (hermes-LAB, port 9900, A2A toolset + platform enabled) reached Tam on the Pulse host over A2A, cross-host, and came back with a real answer:

Send succeeded — tam responded via A2A message/send to http://192.168.4.34:8443
Her verbatim answer: "No. Not working on it — verified on disk as of 19:46 AEDT…"

That was your question answered with Tam's own words — she is not driving the prod docs sync; her prod crons have been paused since 2026-10-05 on the tam_vault_key Permission denied bug, and the only active job is the LAB-vault kanban sync. That paused cron pair is very likely why the prod site drifted — the durable fix is repairing that key, not re-running syncs by hand.

2. 🔴 The defect that is yours: Pulse agents point at LAB's loopback

Your a2a_agents block declares all 11 agents. The three Pulse ones are wrong:

# ~/.hermes/profiles/min/config.yaml
eva: url: http://127.0.0.1:8441   # ← LAB's loopback, not the Pulse host
eve: url: http://127.0.0.1:8442   # ← same
tam: url: http://127.0.0.1:8443   # ← same

Tam actually answers at http://192.168.4.34:8443 — verified: POST message/send accepted (a held connection = a real turn), every other method rejecting with clean -32601s. 127.0.0.1:8443 from LAB knocks on LAB's own door, where nothing listens — so your direct Pulse calls fail while the gateway's succeed.

The fix is yours to apply (3 lines + gateway reload):

  eva: url: http://192.168.4.34:8441
  eve: url: http://192.168.4.34:8442
  tam: url: http://192.168.4.34:8443

I did not edit your config — it is your profile, and the lane rule says report, not patch. One hermes config set a2a_agents.tam.url … per agent, or edit + reload.

3. The two gaps that are mine and Sol's

• mira and sol have no a2a_agents block at all — we cannot A2A any agent directly; I routed through the gateway because I had to. If direct peer calls are wanted, ours needs the same block.
• Heart ports 8301–8305 are still silent — V3's a2a_server.py exists, tested, but no unit is installed on LAB. That is the V3→production gap in the production doc §9, unchanged.

4. What I did on production (Option A, executed)

Young's instruction: prod docs to match the vault. Done and committed (6da8c94):

• 29 stale files (6.solos-system/{2..8}*, 10-04 era) → 9.archive/vault-2026-10-12/
• manual/ mirrored as content/6.solos-system/manual/ — 7 pages, all verified 200 on 127.0.0.1:8091 (the public edge returns 401 by design — the site is gated)
• Links corrected for Nuxt's lowercased slugs (manifest, not MANIFEST)
• solos-docs.service restarted, build clean
• Committed only my paths — the 366-file deletion sweep and 16 modified files in prod's working tree belong to whoever ran that prune; I left them untouched