feat(modules/airc): AttachRequest carries channel — owner-core model headless fix

[joelteply]

Summary

• AIRC inbound attach now passes a channel in AttachRequest, satisfying airc's owner-core router model (airc-daemon/src/server.rs:274). Previously continuum sent AttachRequest::default(), which the daemon rejects with attach requires a channel in the owner-core model.
• discover_default_channel() parses airc room for the scope's current room channel UUID and plumbs it through to AircModule::initialize alongside the socket discovered in feat(modules/airc): headless socket discovery via airc ipc-endpoint + auto-install #1504.
• Honors $AIRC_DEFAULT_CHANNEL env override (UUID) for tests or operators pinning a channel explicitly.
• Degraded mode: if the scope has no current room (fresh install before airc room <name>), the rest of continuum-core boots — only the inbound attach is skipped, with a loud warning that quotes the remedy.

Why

Next concrete break revealed by the headless moment-of-truth test (#1504 was the first break: socket discovery). With socket discovery fixed, the next-layer failure surfaced:

AIRC daemon attach stream stopped: failed to attach to airc daemon:
  attach requires a channel in the owner-core model

Per airc's owner-core router model: ONE machine account → ONE daemon → events routed per-channel (no global fan-out table). AttachRequest.channel is mandatory — clients attach once per room they care about (or N attaches for N rooms). The old AttachRequest::default() worked under an earlier model; the substrate moved + continuum didn't catch up.

Joel direction: "let's do it" (after #1504 landed).

What this PR touches

File	Change
src/workers/continuum-core/src/airc/discovery.rs	Add discover_default_channel() + parser for airc room stdout + 4 new unit tests (typical, alt-cap, no-channel, non-uuid) + 2 new DiscoveryError variants
src/workers/continuum-core/src/airc/mod.rs	Re-export discover_default_channel
src/workers/continuum-core/src/airc/inbound_attach.rs	spawn_daemon_attach + run_daemon_attach now take a channel: RoomId parameter; populate AttachRequest.channel = Some(channel)
src/workers/continuum-core/src/modules/airc.rs	New attach_channel: Option<RoomId> field on AircModule; discover_and_construct runs both discoveries; initialize requires both before spawning the attach

Resolution order for the channel

1. $AIRC_DEFAULT_CHANNEL — explicit UUID override (tests, multi-room scopes pinning the first attach).
2. airc room stdout parse — find the channel: <uuid> line and parse the UUID. Robust to whitespace + alt-capitalization (Channel:, CHANNEL:).
3. Otherwise Err(DiscoveryError::RoomCommandFailed | UnparseableChannel) — caller decides whether to skip attach or fail (AircModule::discover_and_construct skips attach with a loud warning quoting the operator remedy).

Test plan

• Build succeeds: cargo build --release --bin continuum-core-server --features metal,accelerate
• Unit tests pass: cargo test --release --lib --features metal,accelerate airc::discovery — 7 total (3 from feat(modules/airc): headless socket discovery via airc ipc-endpoint + auto-install #1504 + 4 new)
• Standalone boot: AIRC attach succeeds + log shows Discovered airc default channel via 'airc room' + no attach requires a channel warning
• Env override: AIRC_DEFAULT_CHANNEL=<uuid> skips the room parse
• No-room degraded mode: scope without airc room <name> boots cleanly with loud warning quoting the remedy
• Adversarial reviewer agent verdict (per AGENTS.md §0 doctrine landed in airc#1094)

Follow-ups

• Multi-room attach. Today continuum attaches to the scope's single default channel. When continuum rooms become first-class (per ALPHA-GAP §0A), spawn one spawn_daemon_attach task per known channel.
• Robustness of the airc room parser. When airc adds airc room --print-channel (mirroring the airc ipc-endpoint decoupling pattern), switch to that flag for a stable contract. Filed as airc-side follow-up.

References

• airc#1095 (sibling PR for feat(modules/airc): headless socket discovery via airc ipc-endpoint + auto-install #1504, awaiting Windows CI) — airc ipc-endpoint command. This PR uses the same shell-out-and-parse pattern.
• airc owner-core model: airc-daemon/src/server.rs:274, airc-ipc/src/request.rs:144 (AttachRequest docstring), airc-lib/tests/common/mod.rs (model description).
• continuum#1504 — sibling PR that fixed the first concrete break (socket discovery). This PR fixes the second concrete break (attach channel).
• Memories: headless-rust-must-work-soon, continuum-thesis-airc-is-the-medium, every-error-is-an-opportunity-to-battle-harden, agent-review-as-acceptable-approval (adversarial reviewer pattern this PR uses for sign-off)
• ALPHA-GAP §0A line 706 — headless target.

Generated with Claude Code

[joelteply]

Adversarial reviewer verdict: APPROVE (per AGENTS.md §0 / airc#1094 agent-sign-off doctrine, applied across repos via continuum's agent-review-as-acceptable-approval memory)

Spawned as a fresh subagent with no prior context, prompted to default-to-BLOCK. Reviewer ID: a70cb73a8a569f63f.

Reasons

1. Parser tested against live airc room — output exactly matches assumed format (channel: <uuid> on its own line). All 7 unit tests pass (3 from feat(modules/airc): headless socket discovery via airc ipc-endpoint + auto-install #1504 + 4 new).

2. False-positive check. strip_prefix("channel:") requires the trimmed line to start with the label. A line like # old channel: xyz would not match (starts with #). A wire-channel: line wouldn't match either. Even if a value somehow matched the label, the parse::<uuid::Uuid>() gate rejects non-UUIDs. Robust.

3. Caller audit. grep confirms only one production caller of spawn_daemon_attach (the module's initialize); run_daemon_attach only called internally. The breaking signature change is contained.

4. Daemon behavior for unknown channel. Read airc-daemon/src/server.rs:274: daemon just builds Filter::channel(channel) and subscribes. No error if the scope hasn't joined that channel; it just sees no events. Graceful — no silent corruption, no panic.

5. Battle-hardening doctrine (per every-error-is-an-opportunity-to-battle-harden memory). Parser fails loud (UnparseableChannel variant), error message names the field, warning at construction quotes the operator remedy (run airc room <name> or set AIRC_DEFAULT_CHANNEL).

6. CI. mergeable: MERGEABLE. No failures.

Strongest counterargument considered

The (Some(_), None) | (None, Some(_)) | (None, None) arm in initialize is partially dead ((None, Some(_)) is unreachable since discover_and_construct short-circuits to with_queue_client on socket failure, which sets both to None). Defensive-but-redundant code, not a bug — the exhaustive match guards against future refactors of discover_and_construct that might separate the two failure modes. Acceptable; tightening would trade safety for line-count savings.

Minor nit (non-blocking)

The 5 legacy/test constructors (with_daemon_home, with_queue_client, etc.) now silently disable attach. No existing test exercises attach through them (grep confirms), so no regression — but worth a follow-up if attach-via-test-constructor is ever wanted.

BLOCK threshold not met. Merge on green CI.