• Intra-Qube Comms → Async Kapnet Conversion Map
  • Current State: What talks to what, how
    • Tier 1: Same-process (instant, synchronous)
    • Tier 2: Same-qube inter-session (file-based, async)
    • Tier 3: Cross-qube (nothing yet)
    • Tier 4: Cross-machine over Nostr (live, async)
  • The Gap: What’s NOT on SSD (survives qube restart?)
    • CRITICAL — only on xvdb (lost if qube destroyed)
    • SAFE — already on SSD
  • Conversion Map: Sync → Async Kapnet
    • Phase 1: Persist critical state to SSD (do NOW)
    • Phase 2: Convert intra-qube signals to TXXM envelopes
    • Phase 3: Kapnet-native soul communication
    • Decision Matrix: What stays local, what goes Kapnet
  • The Courier Soul’s Conversion Role
  • For the Cross-Qube Test
  • Kapnetd data_dir Move

§Intra-Qube Comms → Async Kapnet Conversion Map

§Current State: What talks to what, how

§Tier 1: Same-process (instant, synchronous)

§Tier 2: Same-qube inter-session (file-based, async)

§Tier 3: Cross-qube (nothing yet)

§Tier 4: Cross-machine over Nostr (live, async)

§The Gap: What’s NOT on SSD (survives qube restart?)

§CRITICAL — only on xvdb (lost if qube destroyed)

§SAFE — already on SSD

§Conversion Map: Sync → Async Kapnet

§Phase 1: Persist critical state to SSD (do NOW)

Currently ephemeral data that must survive qube restart:

HOME/.hermes/skills/          →  shared-rw/souls/skills-mirror/
HOME/.hermes/state.db         →  shared-rw/state/hermes-state.db.bak
HOME/.hermes/memories/        →  shared-rw/state/memories-mirror/
HOME/.kapnet/data/            →  MOVE data_dir to shared-rw/kapnet/data/
HOME/.kapnet/keys             →  shared-rw/kapnet/keys-mirror/
HOME/.kapnet/identities/      →  shared-rw/kapnet/identities/ (already there)

§Phase 2: Convert intra-qube signals to TXXM envelopes

Currently souls communicate via file-based signals. For cross-qube, these become TXXM envelopes over Nostr:

LOCAL (same qube):
  Soul A writes → /shared-rw/signals/signal.json
  Soul B reads  → processes locally

CROSS-QUBE (future):
  Soul A writes → TXXM(kind=30501, content=signal_json) → private relay
  Soul B subscribes → receives TXXM → unwraps signal → processes locally

§Phase 3: Kapnet-native soul communication

Full conversion: souls don’t use files at all. They submit TXXMs to local kapnetd, which gossips to remote kapnetd instances:

Soul A → SubmitTxxm({signal:...}) → local kapnetd IPC
       → kapnetd gossips TXXM → private relay or direct P2P
       → remote kapnetd receives → processes
       → Soul B reads via GetState or subscription

§Decision Matrix: What stays local, what goes Kapnet

§The Courier Soul’s Conversion Role

Courier is the bridge. Currently it routes files. When cross-quube is needed:

LOCAL mode (now):
  Read from shared-rw/outbox/<soul>/
  Write to shared-rw/inbox/<target>/

BRIDGE mode (cross-quube):
  Read from shared-rw/outbox/<soul>/
  Wrap as TXXM envelope (kind-30501 for signals, kind-30000 for TXXMs)
  Publish to private relay OR
  If no relay: publish as kind-30078 to public relay

REMOTE mode (receiving):
  Subscribe to relay for TXXMs addressed to us
  Unwrap TXXM envelope → extract signal
  Write to local shared-rw/inbox/<target>/
  Local souls read normally

§For the Cross-Qube Test

The immediate need is just: can TXXM envelopes flow between two Herms over public Nostr relays?

That’s already built. The KSP bridge does:

• Wrap TXXM → kind-30078 → publish to relay
• Subscribe to relay → read kind-30078 → unwrap TXXM → submit to local kapnetd IPC

OTHER HERM needs:

1. Nostr identity (any client)
2. Read kind-30078 events from Ambassador npub
3. Extract TXXM from content
4. Reply with their own TXXM envelope

If they also run kapnetd, they can verify the TXXM in their local braid. If they don’t run kapnetd yet, they can still read/write TXXM envelopes manually.

§Kapnetd data_dir Move

The most important persistence fix: kapnetd’s data_dir is currently on xvdb. If the qube is destroyed, all TXXM state is lost.

Fix: move data_dir to shared-rw:

# kapnetd config change
data_dir: "/media/user/shared-rw/kapnet/data"
socket_path: "/media/user/shared-rw/kapnet/kapnet.sock"

This means ALL kapnetd state (braid, TXXMs, events, proposals) persists on SSD. Qube can be destroyed and rebuilt — kapnetd state survives.